{
  "name": "Van Data Team AI Model Benchmark Dataset",
  "url": "https://vandatateam.com/ai-model-benchmarks",
  "citation": "Van Data Team, \"AI Model Benchmarks\", dataset version 2026-11-06, https://vandatateam.com/ai-model-benchmarks",
  "citationRequest": "Please cite and link the page above when you use or republish this data.",
  "generatedAt": "2026-11-07",
  "omittedBenchmarks": [
    {
      "id": "aa-speech-arena",
      "name": "Artificial Analysis Speech Arena",
      "source": "https://artificialanalysis.ai/text-to-speech/leaderboard",
      "license": "Artificial Analysis Terms of Use"
    },
    {
      "id": "aa-wer",
      "name": "Artificial Analysis AA-WER v2",
      "source": "https://artificialanalysis.ai/speech-to-text/non-streaming",
      "license": "Artificial Analysis Terms of Use"
    },
    {
      "id": "aa-wer-streaming",
      "name": "Artificial Analysis AA-WER Streaming",
      "source": "https://artificialanalysis.ai/speech-to-text/streaming",
      "license": "Artificial Analysis Terms of Use"
    },
    {
      "id": "aa-video-t2v",
      "name": "Artificial Analysis Video Arena (text to video)",
      "source": "https://artificialanalysis.ai/video/leaderboard/text-to-video",
      "license": "Artificial Analysis Terms of Use"
    },
    {
      "id": "aa-video-i2v",
      "name": "Artificial Analysis Video Arena (image to video)",
      "source": "https://artificialanalysis.ai/video/leaderboard/image-to-video",
      "license": "Artificial Analysis Terms of Use"
    }
  ],
  "omittedNote": "Scores from these benchmarks are shown on the page, credited to their publisher, but their licence does not allow republishing them here. Read them at the source.",
  "schemaVersion": "2",
  "datasetVersion": "2026-11-06",
  "lastUpdated": "2026-11-06",
  "scope": {
    "releaseDateFrom": "2024-01-01",
    "minModels": 30,
    "language": "en"
  },
  "providers": [
    {
      "id": "openai",
      "name": "OpenAI",
      "website": "https://openai.com/"
    },
    {
      "id": "anthropic",
      "name": "Anthropic",
      "website": "https://www.anthropic.com/"
    },
    {
      "id": "google",
      "name": "Google",
      "website": "https://ai.google.dev/"
    },
    {
      "id": "meta",
      "name": "Meta",
      "website": "https://www.llama.com/"
    },
    {
      "id": "xai",
      "name": "xAI",
      "website": "https://x.ai/"
    },
    {
      "id": "deepseek",
      "name": "DeepSeek",
      "website": "https://www.deepseek.com/"
    },
    {
      "id": "mistral",
      "name": "Mistral AI",
      "website": "https://mistral.ai/"
    },
    {
      "id": "qwen",
      "name": "Alibaba Qwen",
      "website": "https://qwenlm.github.io/"
    },
    {
      "id": "kimi",
      "name": "Moonshot Kimi",
      "website": "https://www.moonshot.ai/"
    },
    {
      "id": "runway",
      "name": "Runway",
      "website": "https://runwayml.com/"
    },
    {
      "id": "elevenlabs",
      "name": "ElevenLabs",
      "website": "https://elevenlabs.io/"
    },
    {
      "id": "stability-ai",
      "name": "Stability AI",
      "website": "https://stability.ai/"
    },
    {
      "id": "black-forest-labs",
      "name": "Black Forest Labs",
      "website": "https://bfl.ai/"
    },
    {
      "id": "zai",
      "name": "Z.ai",
      "website": "https://z.ai/"
    },
    {
      "id": "xiaomi",
      "name": "Xiaomi MiMo",
      "website": "https://platform.xiaomimimo.com/"
    },
    {
      "id": "nvidia",
      "name": "NVIDIA",
      "website": "https://www.nvidia.com/"
    },
    {
      "id": "ibm",
      "name": "IBM",
      "website": "https://www.ibm.com/granite"
    },
    {
      "id": "inception",
      "name": "Inception",
      "website": "https://www.inceptionlabs.ai/"
    },
    {
      "id": "upstage",
      "name": "Upstage",
      "website": "https://www.upstage.ai/"
    },
    {
      "id": "stepfun",
      "name": "StepFun",
      "website": "https://platform.stepfun.ai/"
    },
    {
      "id": "microsoft",
      "name": "Microsoft AI",
      "website": "https://microsoft.ai/"
    },
    {
      "id": "cartesia",
      "name": "Cartesia",
      "website": "https://cartesia.ai/"
    },
    {
      "id": "bytedance",
      "name": "ByteDance Seed",
      "website": "https://seed.bytedance.com/"
    },
    {
      "id": "reflection",
      "name": "Reflection AI",
      "website": "https://reflection.ai/"
    },
    {
      "id": "minimax",
      "name": "MiniMax",
      "website": "https://www.minimax.io/"
    },
    {
      "id": "elorian",
      "name": "Elorian",
      "website": "https://elorian.ai/"
    },
    {
      "id": "meituan",
      "name": "Meituan",
      "website": "https://longcat.ai/"
    },
    {
      "id": "assemblyai",
      "name": "AssemblyAI",
      "website": "https://www.assemblyai.com/"
    },
    {
      "id": "tencent",
      "name": "Tencent",
      "website": "https://hy.tencent.ai/"
    },
    {
      "id": "thinking-machines",
      "name": "Thinking Machines Lab",
      "website": "https://thinkingmachines.ai/"
    }
  ],
  "benchmarkDefinitions": [
    {
      "id": "livebench-web-of-lies-v2-2024-11-25",
      "name": "LiveBench Web of Lies V2",
      "category": "reasoning",
      "taskFamily": "text-nlp",
      "description": "Direct accuracy on the Web of Lies V2 reasoning task.",
      "unit": "% correct",
      "direction": "higher-is-better",
      "minScore": 0,
      "maxScore": 100,
      "sourceId": "livebench-2024-11-25-methodology"
    },
    {
      "id": "livebench-overall-2026-06-25",
      "name": "LiveBench Overall",
      "category": "reasoning",
      "taskFamily": "text-nlp",
      "description": "Category-weighted overall score across the seven LiveBench categories in the pinned 2026-06-25 snapshot.",
      "unit": "points",
      "direction": "higher-is-better",
      "minScore": 0,
      "maxScore": 100,
      "sourceId": "livebench-2026-06-25-averaging-19e766a5"
    },
    {
      "id": "epoch-gpqa-diamond",
      "name": "GPQA Diamond",
      "category": "knowledge",
      "taskFamily": "text-nlp",
      "description": "Accuracy on 198 graduate-level multiple-choice science questions, run by Epoch AI. Random guessing scores 25%.",
      "unit": "percent",
      "direction": "higher-is-better",
      "minScore": 0,
      "maxScore": 100,
      "sourceId": "epoch-gpqa-diamond"
    },
    {
      "id": "epoch-frontiermath-tiers-1-3-v2",
      "name": "FrontierMath Tiers 1-3 (v2)",
      "category": "reasoning",
      "taskFamily": "text-nlp",
      "description": "Share of 295 private research-level math problems solved, run by Epoch AI on the v2 problem set.",
      "unit": "percent",
      "direction": "higher-is-better",
      "minScore": 0,
      "maxScore": 100,
      "sourceId": "epoch-frontiermath"
    },
    {
      "id": "epoch-simpleqa-verified",
      "name": "SimpleQA Verified",
      "category": "knowledge",
      "taskFamily": "text-nlp",
      "description": "Share of 1,000 short factual questions answered correctly, run by Epoch AI. Higher means fewer confident factual errors.",
      "unit": "percent",
      "direction": "higher-is-better",
      "minScore": 0,
      "maxScore": 100,
      "sourceId": "epoch-simpleqa-verified"
    },
    {
      "id": "arena-text-to-image",
      "name": "Arena Text-to-Image",
      "category": "multimodal",
      "taskFamily": "image",
      "description": "Human-preference rating from blind pairwise votes on Arena, where people compare images generated from the same prompt and pick the better one. A 100-point gap means the higher-rated model wins about 64% of comparisons.",
      "unit": "Elo",
      "direction": "higher-is-better",
      "minScore": 800,
      "maxScore": 1600,
      "sourceId": "arena-text-to-image"
    },
    {
      "id": "arena-image-edit",
      "name": "Arena Image Edit",
      "category": "multimodal",
      "taskFamily": "image",
      "description": "Human-preference rating from blind pairwise votes on Arena, where people compare edits of the same image from the same instruction and pick the better one. A 100-point gap means the higher-rated model wins about 64% of comparisons.",
      "unit": "Elo",
      "direction": "higher-is-better",
      "minScore": 800,
      "maxScore": 1600,
      "sourceId": "arena-image-edit"
    },
    {
      "id": "arena-text-to-video",
      "name": "Arena Text-to-Video",
      "category": "multimodal",
      "taskFamily": "video",
      "description": "Human-preference rating from blind pairwise votes on Arena, where people compare videos generated from the same prompt and pick the better one. A 100-point gap means the higher-rated model wins about 64% of comparisons.",
      "unit": "Elo",
      "direction": "higher-is-better",
      "minScore": 800,
      "maxScore": 1600,
      "sourceId": "arena-text-to-video"
    },
    {
      "id": "arena-image-to-video",
      "name": "Arena Image-to-Video",
      "category": "multimodal",
      "taskFamily": "video",
      "description": "Human-preference rating from blind pairwise votes on Arena, where people compare videos animated from the same image and prompt and pick the better one. A 100-point gap means the higher-rated model wins about 64% of comparisons.",
      "unit": "Elo",
      "direction": "higher-is-better",
      "minScore": 800,
      "maxScore": 1600,
      "sourceId": "arena-image-to-video"
    },
    {
      "id": "arena-text",
      "name": "Arena Text",
      "category": "reasoning",
      "taskFamily": "text-nlp",
      "description": "Human-preference rating from blind pairwise votes on Arena, where people compare two anonymous models on the same prompt and pick the better answer. Style-controlled, as on arena.ai. A 100-point gap means the higher-rated model wins about 64% of comparisons.",
      "unit": "Elo",
      "direction": "higher-is-better",
      "minScore": 800,
      "maxScore": 1600,
      "sourceId": "arena-text"
    },
    {
      "id": "arena-webdev",
      "name": "Arena WebDev",
      "category": "coding",
      "taskFamily": "text-nlp",
      "description": "Human-preference rating from blind pairwise votes in Code Arena on arena.ai, where people ask two anonymous models to build the same web app, try both and pick the better one. A 100-point gap means the higher-rated model wins about 64% of comparisons.",
      "unit": "Elo",
      "direction": "higher-is-better",
      "minScore": 800,
      "maxScore": 2000,
      "sourceId": "arena-webdev"
    }
  ],
  "benchmarkSnapshots": [
    {
      "id": "livebench-web-of-lies-v2-2024-11-25-snapshot-2024-11-25",
      "benchmarkId": "livebench-web-of-lies-v2-2024-11-25",
      "taskFamily": "text-nlp",
      "version": "2024-11-25",
      "measuredAt": "2024-11-25",
      "sourceId": "livebench-2024-11-25-results-347e3d0b"
    },
    {
      "id": "livebench-overall-2026-06-25-snapshot-2026-06-25",
      "benchmarkId": "livebench-overall-2026-06-25",
      "taskFamily": "text-nlp",
      "version": "2026-06-25",
      "measuredAt": "2026-06-25",
      "sourceId": "livebench-2026-06-25-results-19e766a5"
    },
    {
      "id": "epoch-gpqa-diamond-snapshot-2026-11-03",
      "benchmarkId": "epoch-gpqa-diamond",
      "taskFamily": "text-nlp",
      "version": "Epoch AI benchmark data 2026-11-03",
      "measuredAt": "2026-11-03",
      "sourceId": "epoch-benchmark-data-2026-11-03"
    },
    {
      "id": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
      "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
      "taskFamily": "text-nlp",
      "version": "Epoch AI benchmark data 2026-11-03",
      "measuredAt": "2026-11-03",
      "sourceId": "epoch-benchmark-data-2026-11-03"
    },
    {
      "id": "epoch-simpleqa-verified-snapshot-2026-11-03",
      "benchmarkId": "epoch-simpleqa-verified",
      "taskFamily": "text-nlp",
      "version": "Epoch AI benchmark data 2026-11-03",
      "measuredAt": "2026-11-03",
      "sourceId": "epoch-benchmark-data-2026-11-03"
    },
    {
      "id": "arena-text-snapshot-2026-10-28",
      "benchmarkId": "arena-text",
      "taskFamily": "text-nlp",
      "version": "Arena leaderboard published 2026-10-28",
      "measuredAt": "2026-10-28",
      "sourceId": "arena-leaderboard-dataset-2026-11-03"
    },
    {
      "id": "arena-webdev-snapshot-2026-10-28",
      "benchmarkId": "arena-webdev",
      "taskFamily": "text-nlp",
      "version": "Arena leaderboard published 2026-10-28",
      "measuredAt": "2026-10-28",
      "sourceId": "arena-leaderboard-dataset-2026-11-03"
    },
    {
      "id": "arena-text-to-image-snapshot-2026-11-04",
      "benchmarkId": "arena-text-to-image",
      "taskFamily": "image",
      "version": "Arena leaderboard published 2026-11-04",
      "measuredAt": "2026-11-04",
      "sourceId": "arena-leaderboard-dataset-2026-11-06"
    },
    {
      "id": "arena-image-edit-snapshot-2026-11-04",
      "benchmarkId": "arena-image-edit",
      "taskFamily": "image",
      "version": "Arena leaderboard published 2026-11-04",
      "measuredAt": "2026-11-04",
      "sourceId": "arena-leaderboard-dataset-2026-11-06"
    },
    {
      "id": "arena-text-to-video-snapshot-2026-11-05",
      "benchmarkId": "arena-text-to-video",
      "taskFamily": "video",
      "version": "Arena leaderboard published 2026-11-05",
      "measuredAt": "2026-11-05",
      "sourceId": "arena-leaderboard-dataset-2026-11-06"
    },
    {
      "id": "arena-image-to-video-snapshot-2026-11-05",
      "benchmarkId": "arena-image-to-video",
      "taskFamily": "video",
      "version": "Arena leaderboard published 2026-11-05",
      "measuredAt": "2026-11-05",
      "sourceId": "arena-leaderboard-dataset-2026-11-06"
    }
  ],
  "models": [
    {
      "id": "gpt-4o",
      "providerId": "openai",
      "name": "GPT-4o",
      "family": "GPT-4o",
      "releaseDate": "2024-05-13",
      "status": "legacy",
      "access": "api",
      "contextWindowTokens": 128000,
      "maxOutputTokens": 16384,
      "modalities": [
        "text",
        "image",
        "audio"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "GPT-4o records API access for text, image, and audio inputs and a published 128,000-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "API trials combining text, images, and audio"
      ],
      "pricing": {
        "inputPerMillionUsd": 2.5,
        "outputPerMillionUsd": 10,
        "cachedInputPerMillionUsd": 1.25,
        "sourceId": "openai-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2.5,
          "tier": "standard",
          "sourceId": "openai-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 10,
          "tier": "standard",
          "sourceId": "openai-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 1.25,
          "tier": "cached",
          "sourceId": "openai-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 48.9,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-4o-2024-05-13"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1346,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-4o (2024-05-13)"
        }
      ],
      "officialSourceIds": [
        "openai-gpt-4o-release",
        "openai-pricing",
        "openai-gpt-4o-model"
      ]
    },
    {
      "id": "gpt-41",
      "providerId": "openai",
      "name": "GPT-4.1",
      "family": "GPT-4.1",
      "releaseDate": "2025-04-14",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1047576,
      "maxOutputTokens": 32768,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "GPT-4.1 records API access for text and image inputs and a published 1,047,576-token context window"
      ],
      "cautions": [
        "Not in the pinned LiveBench snapshot; its independent scores here come from Epoch AI's own GPQA Diamond, FrontierMath and SimpleQA Verified runs"
      ],
      "bestFor": [
        "Long-context API trials using text and image inputs"
      ],
      "pricing": {
        "inputPerMillionUsd": 2,
        "outputPerMillionUsd": 8,
        "cachedInputPerMillionUsd": 0.5,
        "sourceId": "openai-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "standard",
          "sourceId": "openai-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 8,
          "tier": "standard",
          "sourceId": "openai-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.5,
          "tier": "cached",
          "sourceId": "openai-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 66.9,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-4.1-2025-04-14"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-4.1-2025-04-14"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 31.1,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-4.1-2025-04-14"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1415,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-4.1 (2025-04-14)"
        }
      ],
      "officialSourceIds": [
        "openai-gpt-4-1-release",
        "openai-pricing",
        "openai-gpt-4-1-model"
      ]
    },
    {
      "id": "gpt-41-mini",
      "providerId": "openai",
      "name": "GPT-4.1 mini",
      "family": "GPT-4.1",
      "releaseDate": "2025-04-14",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1047576,
      "maxOutputTokens": 32768,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "GPT-4.1 mini records API access for text and image inputs and a published 1,047,576-token context window"
      ],
      "cautions": [
        "Not in the pinned LiveBench snapshot; its independent scores here come from Epoch AI's own GPQA Diamond, FrontierMath and SimpleQA Verified runs"
      ],
      "bestFor": [
        "Cost-sensitive long-context API evaluation"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.4,
        "outputPerMillionUsd": 1.6,
        "cachedInputPerMillionUsd": 0.1,
        "sourceId": "openai-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.4,
          "tier": "standard",
          "sourceId": "openai-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 1.6,
          "tier": "standard",
          "sourceId": "openai-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.1,
          "tier": "cached",
          "sourceId": "openai-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 65.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-4.1-mini-2025-04-14"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 6.7,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-4.1-mini-2025-04-14"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 12.7,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-4.1-mini-2025-04-14"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1383,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-4.1 mini (2025-04-14)"
        }
      ],
      "officialSourceIds": [
        "openai-gpt-4-1-release",
        "openai-pricing",
        "openai-gpt-4-1-mini-model"
      ]
    },
    {
      "id": "gpt-41-nano",
      "providerId": "openai",
      "name": "GPT-4.1 nano",
      "family": "GPT-4.1",
      "releaseDate": "2025-04-14",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1047576,
      "maxOutputTokens": 32768,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "GPT-4.1 nano records API access for text and image inputs and a published 1,047,576-token context window"
      ],
      "cautions": [
        "Not in the pinned LiveBench snapshot; its independent scores here come from Epoch AI's own GPQA Diamond, FrontierMath and SimpleQA Verified runs"
      ],
      "bestFor": [
        "High-volume API screening at low published token prices"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.1,
        "outputPerMillionUsd": 0.4,
        "cachedInputPerMillionUsd": 0.025,
        "sourceId": "openai-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.1,
          "tier": "standard",
          "sourceId": "openai-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 0.4,
          "tier": "standard",
          "sourceId": "openai-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.025,
          "tier": "cached",
          "sourceId": "openai-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 48.9,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-4.1-nano-2025-04-14"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-4.1-nano-2025-04-14"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1322,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-4.1 nano (2025-04-14)"
        }
      ],
      "officialSourceIds": [
        "openai-gpt-4-1-release",
        "openai-pricing",
        "openai-gpt-4-1-nano-model"
      ]
    },
    {
      "id": "claude-3-opus",
      "providerId": "anthropic",
      "name": "Claude 3 Opus",
      "family": "Claude 3",
      "releaseDate": "2024-03-04",
      "status": "legacy",
      "access": "api",
      "contextWindowTokens": 200000,
      "maxOutputTokens": 4096,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Claude 3 Opus records API access for text and image inputs and a published 200,000-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical comparison for systems that still use Claude 3 Opus"
      ],
      "pricing": {
        "inputPerMillionUsd": 15,
        "outputPerMillionUsd": 75,
        "cachedInputPerMillionUsd": 1.5,
        "sourceId": "anthropic-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 15,
          "tier": "standard",
          "sourceId": "anthropic-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 75,
          "tier": "standard",
          "sourceId": "anthropic-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 1.5,
          "tier": "cached",
          "sourceId": "anthropic-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "livebench-web-of-lies-v2-2024-11-25",
          "benchmarkSnapshotId": "livebench-web-of-lies-v2-2024-11-25-snapshot-2024-11-25",
          "score": 50,
          "sourceId": "livebench-2024-11-25-results-347e3d0b",
          "measuredAt": "2024-11-25",
          "sourceRow": "claude-3-opus-20240229"
        },
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 47.2,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-3-opus-20240229"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 12.6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-3-opus-20240229"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1322,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude 3 Opus (20240229)"
        }
      ],
      "officialSourceIds": [
        "anthropic-claude-3-release",
        "anthropic-pricing",
        "anthropic-models-overview"
      ]
    },
    {
      "id": "claude-3-sonnet",
      "providerId": "anthropic",
      "name": "Claude 3 Sonnet",
      "family": "Claude 3",
      "releaseDate": "2024-03-04",
      "status": "legacy",
      "access": "api",
      "contextWindowTokens": 200000,
      "maxOutputTokens": 4096,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Claude 3 Sonnet records API access for text and image inputs and a published 200,000-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical text-and-image API comparison at mid-tier list pricing"
      ],
      "pricing": {
        "inputPerMillionUsd": 3,
        "outputPerMillionUsd": 15,
        "cachedInputPerMillionUsd": null,
        "sourceId": "anthropic-claude-3-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 3,
          "tier": "standard",
          "sourceId": "anthropic-claude-3-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": 15,
          "tier": "standard",
          "sourceId": "anthropic-claude-3-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "anthropic-claude-3-release"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 40.6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-3-sonnet-20240229"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1281,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude 3 Sonnet (20240229)"
        }
      ],
      "officialSourceIds": [
        "anthropic-claude-3-release",
        "anthropic-pricing",
        "anthropic-models-overview"
      ]
    },
    {
      "id": "claude-3-haiku",
      "providerId": "anthropic",
      "name": "Claude 3 Haiku",
      "family": "Claude 3",
      "releaseDate": "2024-03-13",
      "status": "legacy",
      "access": "api",
      "contextWindowTokens": 200000,
      "maxOutputTokens": 4096,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Claude 3 Haiku records API access for text and image inputs and a published 200,000-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical cost-sensitive text-and-image API comparison"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.25,
        "outputPerMillionUsd": 1.25,
        "cachedInputPerMillionUsd": 0.03,
        "sourceId": "anthropic-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.25,
          "tier": "standard",
          "sourceId": "anthropic-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 1.25,
          "tier": "standard",
          "sourceId": "anthropic-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.03,
          "tier": "cached",
          "sourceId": "anthropic-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 36.3,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-3-haiku-20240307"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1262,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude 3 Haiku (20240307)"
        }
      ],
      "officialSourceIds": [
        "anthropic-claude-3-haiku-release",
        "anthropic-models-overview",
        "anthropic-pricing"
      ]
    },
    {
      "id": "claude-3-7-sonnet",
      "providerId": "anthropic",
      "name": "Claude 3.7 Sonnet",
      "family": "Claude 3.7",
      "releaseDate": "2025-02-24",
      "status": "legacy",
      "access": "api",
      "contextWindowTokens": 200000,
      "maxOutputTokens": 64000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Claude 3.7 Sonnet records API access for text and image inputs and a published 200,000-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Reasoning screening that may require long generated responses"
      ],
      "pricing": {
        "inputPerMillionUsd": 3,
        "outputPerMillionUsd": 15,
        "cachedInputPerMillionUsd": 0.3,
        "sourceId": "anthropic-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 3,
          "tier": "standard",
          "sourceId": "anthropic-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 15,
          "tier": "standard",
          "sourceId": "anthropic-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.3,
          "tier": "cached",
          "sourceId": "anthropic-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "livebench-web-of-lies-v2-2024-11-25",
          "benchmarkSnapshotId": "livebench-web-of-lies-v2-2024-11-25-snapshot-2024-11-25",
          "score": 94,
          "sourceId": "livebench-2024-11-25-results-347e3d0b",
          "measuredAt": "2024-11-25",
          "sourceRow": "claude-3-7-sonnet"
        },
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 79.7,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-3-7-sonnet-20250219_64K"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1388,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude 3.7 Sonnet (Thinking 32K, 20250219)"
        }
      ],
      "officialSourceIds": [
        "anthropic-claude-3-7-release",
        "anthropic-pricing",
        "anthropic-models-overview"
      ]
    },
    {
      "id": "gemini-1-5-pro",
      "providerId": "google",
      "name": "Gemini 1.5 Pro",
      "family": "Gemini 1.5",
      "releaseDate": "2024-02-15",
      "status": "legacy",
      "access": "api",
      "contextWindowTokens": 2097152,
      "maxOutputTokens": 8192,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Gemini 1.5 Pro records API access for text, image, audio, and video inputs and a published 2,097,152-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical very-long-context multimodal API comparison"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 45.9,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-1.5-pro-001"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1324,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 1.5 Pro (001)"
        }
      ],
      "officialSourceIds": [
        "google-gemini-1-5-release",
        "google-pricing",
        "google-models-overview"
      ]
    },
    {
      "id": "gemini-1-5-flash",
      "providerId": "google",
      "name": "Gemini 1.5 Flash",
      "family": "Gemini 1.5",
      "releaseDate": "2024-05-14",
      "status": "legacy",
      "access": "api",
      "contextWindowTokens": 1048576,
      "maxOutputTokens": 8192,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Gemini 1.5 Flash records API access for text, image, audio, and video inputs and a published 1,048,576-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical multimodal API comparison with long context"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 40.4,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-1.5-flash-001"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1287,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 1.5 Flash (001)"
        }
      ],
      "officialSourceIds": [
        "google-gemini-1-5-flash-release",
        "google-pricing",
        "google-models-overview"
      ]
    },
    {
      "id": "gemini-2-0-flash",
      "providerId": "google",
      "name": "Gemini 2.0 Flash",
      "family": "Gemini 2.0",
      "releaseDate": "2024-12-11",
      "status": "legacy",
      "access": "api",
      "contextWindowTokens": 1048576,
      "maxOutputTokens": 8192,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Gemini 2.0 Flash records API access for text, image, audio, and video inputs and a published 1,048,576-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical multimodal API reasoning comparison"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "livebench-web-of-lies-v2-2024-11-25",
          "benchmarkSnapshotId": "livebench-web-of-lies-v2-2024-11-25-snapshot-2024-11-25",
          "score": 74,
          "sourceId": "livebench-2024-11-25-results-347e3d0b",
          "measuredAt": "2024-11-25",
          "sourceRow": "gemini-2.0-flash"
        },
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 64.1,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-2.0-flash-001"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1360,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 2.0 Flash (001)"
        }
      ],
      "officialSourceIds": [
        "google-gemini-2-release",
        "google-pricing",
        "google-models-overview"
      ]
    },
    {
      "id": "gemini-2-5-pro",
      "providerId": "google",
      "name": "Gemini 2.5 Pro",
      "family": "Gemini 2.5",
      "releaseDate": "2025-03-25",
      "status": "legacy",
      "deprecation": {
        "date": "2026-11-02",
        "replacementId": "gemini-38-flash",
        "sourceId": "google-gemini-changelog-2026-11-02"
      },
      "access": "api",
      "contextWindowTokens": 1048576,
      "maxOutputTokens": 65536,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Gemini 2.5 Pro records API access for text, image, audio, and video inputs and a published 1,048,576-token context window"
      ],
      "cautions": [
        "Google deprecated it on 2026-11-02: Google AI Studio no longer offers it, and from 2026-12-01 API requests to gemini-2.5-pro are routed to gemini-3.8-flash and billed at its rates",
        "Not in the pinned LiveBench snapshot; its independent scores here come from Epoch AI's own GPQA Diamond, FrontierMath and SimpleQA Verified runs"
      ],
      "bestFor": [
        "Long-context API trials spanning text, image, audio, and video"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 85.3,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-2.5-pro"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 24.6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-2.5-pro"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1445,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 2.5 Pro"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1226,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 2.5 Pro"
        }
      ],
      "officialSourceIds": [
        "google-gemini-2-5-release",
        "google-pricing",
        "google-gemini-2-5-pro-model",
        "google-gemini-changelog-2026-11-02"
      ]
    },
    {
      "id": "llama-3-8b",
      "providerId": "meta",
      "name": "Llama 3 8B",
      "family": "Llama 3",
      "releaseDate": "2024-04-18",
      "status": "legacy",
      "access": "open-weights",
      "contextWindowTokens": 8192,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Llama 3 8B records open-weight access for text inputs and a published 8,192-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical small open-weight text deployment baselines"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "meta-llama-3-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "meta-llama-3-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "meta-llama-3-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "meta-llama-3-release"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1224,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Llama 3 8B Instruct"
        }
      ],
      "officialSourceIds": [
        "meta-llama-3-release"
      ]
    },
    {
      "id": "llama-3-70b",
      "providerId": "meta",
      "name": "Llama 3 70B",
      "family": "Llama 3",
      "releaseDate": "2024-04-18",
      "status": "legacy",
      "access": "open-weights",
      "contextWindowTokens": 8192,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Llama 3 70B records open-weight access for text inputs and a published 8,192-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical larger open-weight text deployment baselines"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "meta-llama-3-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "meta-llama-3-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "meta-llama-3-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "meta-llama-3-release"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1277,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Llama 3 70B Instruct"
        }
      ],
      "officialSourceIds": [
        "meta-llama-3-release"
      ]
    },
    {
      "id": "llama-3-1-405b",
      "providerId": "meta",
      "name": "Llama 3.1 405B",
      "family": "Llama 3.1",
      "releaseDate": "2024-07-23",
      "status": "legacy",
      "access": "open-weights",
      "contextWindowTokens": 128000,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Llama 3.1 405B records open-weight access for text inputs and a published 128,000-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical large open-weight evaluations with 128,000-token context"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "meta-llama-3-1-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "meta-llama-3-1-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "meta-llama-3-1-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "meta-llama-3-1-release"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1335,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Llama 3.1 405B Instruct (BF16)"
        }
      ],
      "officialSourceIds": [
        "meta-llama-3-1-release"
      ]
    },
    {
      "id": "llama-3-2-11b-vision",
      "providerId": "meta",
      "name": "Llama 3.2 11B Vision",
      "family": "Llama 3.2",
      "releaseDate": "2024-09-25",
      "status": "legacy",
      "access": "open-weights",
      "contextWindowTokens": 128000,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Llama 3.2 11B Vision records open-weight access for text and image inputs and a published 128,000-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical private text-and-image deployment evaluation"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "meta-llama-3-2-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "meta-llama-3-2-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "meta-llama-3-2-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "meta-llama-3-2-release"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "meta-llama-3-2-release"
      ]
    },
    {
      "id": "grok-1-5",
      "providerId": "xai",
      "name": "Grok-1.5",
      "family": "Grok 1.5",
      "releaseDate": "2024-03-28",
      "status": "legacy",
      "access": "api",
      "contextWindowTokens": 128000,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Grok-1.5 records API access for text inputs and a published 128,000-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical API evaluations needing a documented 128,000-token context"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "xai-models-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "xai-models-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "xai-models-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "xai-models-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "xai-grok-1-5-release",
        "xai-models-pricing"
      ]
    },
    {
      "id": "grok-1-5v",
      "providerId": "xai",
      "name": "Grok-1.5V",
      "family": "Grok 1.5",
      "releaseDate": "2024-04-12",
      "status": "legacy",
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Grok-1.5V records API access for text and image inputs"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical text-and-image API comparison"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "xai-models-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "xai-models-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "xai-models-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "xai-models-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "xai-grok-1-5v-release",
        "xai-models-pricing"
      ]
    },
    {
      "id": "grok-2",
      "providerId": "xai",
      "name": "Grok 2",
      "family": "Grok 2",
      "releaseDate": "2024-08-13",
      "status": "legacy",
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Grok 2 records API access for text inputs"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical text API comparison where limits can be verified separately"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "xai-models-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "xai-models-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "xai-models-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "xai-models-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1336,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Grok 2 (2024-08-13)"
        }
      ],
      "officialSourceIds": [
        "xai-grok-2-release",
        "xai-models-pricing"
      ]
    },
    {
      "id": "grok-3",
      "providerId": "xai",
      "name": "Grok 3",
      "family": "Grok 3",
      "releaseDate": "2025-02-19",
      "status": "legacy",
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Grok 3 records API access for text inputs"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical text API comparison where limits can be verified separately"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "xai-models-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "xai-models-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "xai-models-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "xai-models-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 75.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "grok-3-beta"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1411,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Grok 3 Preview (02-24)"
        }
      ],
      "officialSourceIds": [
        "xai-grok-3-release",
        "xai-models-pricing",
        "xai-may-15-retirement"
      ]
    },
    {
      "id": "deepseek-v2-chat",
      "providerId": "deepseek",
      "name": "DeepSeek-V2 Chat",
      "family": "DeepSeek-V2",
      "releaseDate": "2024-05-06",
      "status": "legacy",
      "access": "both",
      "contextWindowTokens": 128000,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "DeepSeek-V2 Chat records both API and open-weight access for text inputs and a published 128,000-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical API or open-weight conversational evaluation"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "deepseek-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "deepseek-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "deepseek-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "deepseek-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "deepseek-v2-release",
        "deepseek-pricing"
      ]
    },
    {
      "id": "deepseek-coder-v2-instruct",
      "providerId": "deepseek",
      "name": "DeepSeek-Coder-V2 Instruct",
      "family": "DeepSeek-Coder-V2",
      "releaseDate": "2024-06-17",
      "status": "legacy",
      "access": "both",
      "contextWindowTokens": 128000,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "DeepSeek-Coder-V2 Instruct records both API and open-weight access for text inputs and a published 128,000-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Repository and code-generation trials with API or open-weight deployment"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "deepseek-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "deepseek-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "deepseek-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "deepseek-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1265,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "DeepSeek-Coder-V2"
        }
      ],
      "officialSourceIds": [
        "deepseek-coder-v2-release",
        "deepseek-pricing"
      ]
    },
    {
      "id": "deepseek-v3",
      "providerId": "deepseek",
      "name": "DeepSeek-V3",
      "family": "DeepSeek-V3",
      "releaseDate": "2024-12-26",
      "status": "legacy",
      "access": "both",
      "contextWindowTokens": 128000,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "DeepSeek-V3 records both API and open-weight access for text inputs and a published 128,000-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical reasoning comparison with API or open-weight access"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "deepseek-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "deepseek-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "deepseek-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "deepseek-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "livebench-web-of-lies-v2-2024-11-25",
          "benchmarkSnapshotId": "livebench-web-of-lies-v2-2024-11-25-snapshot-2024-11-25",
          "score": 86,
          "sourceId": "livebench-2024-11-25-results-347e3d0b",
          "measuredAt": "2024-11-25",
          "sourceRow": "deepseek-v3"
        },
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 56.5,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "DeepSeek-V3"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1358,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "DeepSeek-V3"
        }
      ],
      "officialSourceIds": [
        "deepseek-v3-release",
        "deepseek-pricing"
      ]
    },
    {
      "id": "deepseek-r1",
      "providerId": "deepseek",
      "name": "DeepSeek-R1",
      "family": "DeepSeek-R1",
      "releaseDate": "2025-01-20",
      "status": "active",
      "access": "both",
      "contextWindowTokens": 128000,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "DeepSeek-R1 records both API and open-weight access for text inputs and a published 128,000-token context window"
      ],
      "cautions": [
        "Hosted token price and infrastructure cost are not represented; measure deployment and operating cost directly"
      ],
      "bestFor": [
        "Reasoning trials requiring API or open-weight deployment options"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "deepseek-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "deepseek-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "deepseek-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "deepseek-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "livebench-web-of-lies-v2-2024-11-25",
          "benchmarkSnapshotId": "livebench-web-of-lies-v2-2024-11-25-snapshot-2024-11-25",
          "score": 100,
          "sourceId": "livebench-2024-11-25-results-347e3d0b",
          "measuredAt": "2024-11-25",
          "sourceRow": "deepseek-r1"
        },
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 71.7,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "DeepSeek-R1"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1398,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "DeepSeek-R1"
        }
      ],
      "officialSourceIds": [
        "deepseek-r1-release",
        "deepseek-pricing"
      ]
    },
    {
      "id": "codestral-22b",
      "providerId": "mistral",
      "name": "Codestral 22B",
      "family": "Codestral",
      "releaseDate": "2024-05-29",
      "status": "legacy",
      "access": "both",
      "contextWindowTokens": 32000,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Codestral 22B records both API and open-weight access for text inputs and a published 32,000-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Code-generation trials with API or open-weight deployment"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "mistral-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "mistral-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "mistral-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "mistral-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "mistral-codestral-release",
        "mistral-pricing"
      ]
    },
    {
      "id": "mistral-large-2",
      "providerId": "mistral",
      "name": "Mistral Large 2",
      "family": "Mistral Large",
      "releaseDate": "2024-07-24",
      "status": "legacy",
      "access": "both",
      "contextWindowTokens": 128000,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Mistral Large 2 records both API and open-weight access for text inputs and a published 128,000-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical general-purpose text trials with flexible deployment"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "mistral-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "mistral-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "mistral-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "mistral-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 49,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "mistral-large-2407"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1314,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Mistral Large 2 (2407)"
        }
      ],
      "officialSourceIds": [
        "mistral-large-2-release",
        "mistral-pricing"
      ]
    },
    {
      "id": "pixtral-12b",
      "providerId": "mistral",
      "name": "Pixtral 12B",
      "family": "Pixtral",
      "releaseDate": "2024-09-17",
      "status": "legacy",
      "access": "both",
      "contextWindowTokens": 128000,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Pixtral 12B records both API and open-weight access for text and image inputs and a published 128,000-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical private text-and-image evaluation"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "mistral-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "mistral-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "mistral-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "mistral-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "mistral-pixtral-release",
        "mistral-pricing"
      ]
    },
    {
      "id": "ministral-8b",
      "providerId": "mistral",
      "name": "Ministral 8B",
      "family": "Ministral",
      "releaseDate": "2024-10-16",
      "status": "legacy",
      "access": "both",
      "contextWindowTokens": 128000,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Ministral 8B records both API and open-weight access for text inputs and a published 128,000-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical compact open-weight or API text deployments"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "mistral-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "mistral-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "mistral-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "mistral-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 27.1,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "ministral-8b-2410"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1238,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Ministral 8B (2410)"
        }
      ],
      "officialSourceIds": [
        "mistral-ministral-release",
        "mistral-pricing"
      ]
    },
    {
      "id": "qwen2-7b-instruct",
      "providerId": "qwen",
      "name": "Qwen2 7B Instruct",
      "family": "Qwen2",
      "releaseDate": "2024-06-07",
      "status": "legacy",
      "access": "open-weights",
      "contextWindowTokens": 128000,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Qwen2 7B Instruct records open-weight access for text inputs and a published 128,000-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical compact private text deployments"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "qwen2-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "qwen2-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "qwen2-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "qwen2-release"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "qwen2-release"
      ]
    },
    {
      "id": "qwen2-72b-instruct",
      "providerId": "qwen",
      "name": "Qwen2 72B Instruct",
      "family": "Qwen2",
      "releaseDate": "2024-06-07",
      "status": "legacy",
      "access": "open-weights",
      "contextWindowTokens": 128000,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Qwen2 72B Instruct records open-weight access for text inputs and a published 128,000-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical larger private text deployments"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "qwen2-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "qwen2-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "qwen2-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "qwen2-release"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 40.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "qwen2-72b-instruct"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1262,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Qwen2-72B-Instruct"
        }
      ],
      "officialSourceIds": [
        "qwen2-release"
      ]
    },
    {
      "id": "qwen2-5-32b-instruct",
      "providerId": "qwen",
      "name": "Qwen2.5 32B Instruct",
      "family": "Qwen2.5",
      "releaseDate": "2024-09-19",
      "status": "legacy",
      "access": "open-weights",
      "contextWindowTokens": 128000,
      "maxOutputTokens": 8192,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Qwen2.5 32B Instruct records open-weight access for text inputs and a published 128,000-token context window"
      ],
      "cautions": [
        "This record is marked legacy; confirm current availability and migration requirements before adoption"
      ],
      "bestFor": [
        "Historical private text deployments with a published output limit"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "qwen2-5-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "qwen2-5-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "qwen2-5-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "qwen2-5-release"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 46.1,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "qwen2.5-32b-instruct"
        }
      ],
      "officialSourceIds": [
        "qwen2-5-release"
      ]
    },
    {
      "id": "qwen2-5-vl-72b-instruct",
      "providerId": "qwen",
      "name": "Qwen2.5-VL 72B Instruct",
      "family": "Qwen2.5-VL",
      "releaseDate": "2025-01-26",
      "status": "active",
      "access": "open-weights",
      "contextWindowTokens": 128000,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Qwen2.5-VL 72B Instruct records open-weight access for text, image, and video inputs and a published 128,000-token context window"
      ],
      "cautions": [
        "Hosted token price and infrastructure cost are not represented; measure deployment and operating cost directly"
      ],
      "bestFor": [
        "Private text, image, and video evaluation"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "qwen2-5-vl-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "qwen2-5-vl-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "qwen2-5-vl-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "qwen2-5-vl-release"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "qwen2-5-vl-release"
      ]
    },
    {
      "id": "kimi-k2-base",
      "providerId": "kimi",
      "name": "Kimi K2 Base",
      "family": "Kimi K2",
      "releaseDate": "2025-07-11",
      "status": "active",
      "access": "open-weights",
      "contextWindowTokens": 128000,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Kimi K2 Base records open-weight access for text inputs and a published 128,000-token context window"
      ],
      "cautions": [
        "Hosted token price and infrastructure cost are not represented; measure deployment and operating cost directly"
      ],
      "bestFor": [
        "Base-model adaptation with open-weight infrastructure control"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "kimi-k2-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "kimi-k2-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "kimi-k2-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "kimi-k2-release"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "kimi-k2-release"
      ]
    },
    {
      "id": "kimi-vl-a3b-instruct",
      "providerId": "kimi",
      "name": "Kimi-VL-A3B-Instruct",
      "family": "Kimi-VL",
      "releaseDate": "2025-04-09",
      "status": "active",
      "access": "open-weights",
      "contextWindowTokens": 128000,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Kimi-VL-A3B-Instruct records open-weight access for text and image inputs and a published 128,000-token context window"
      ],
      "cautions": [
        "Hosted token price and infrastructure cost are not represented; measure deployment and operating cost directly"
      ],
      "bestFor": [
        "Private text-and-image evaluation"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "kimi-vl-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "kimi-vl-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "kimi-vl-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "kimi-vl-release"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "kimi-vl-release"
      ]
    },
    {
      "id": "kimi-dev-72b",
      "providerId": "kimi",
      "name": "Kimi-Dev-72B",
      "family": "Kimi-Dev",
      "releaseDate": "2025-06-17",
      "status": "active",
      "access": "open-weights",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Kimi-Dev-72B records open-weight access for text inputs"
      ],
      "cautions": [
        "Hosted token price and infrastructure cost are not represented; measure deployment and operating cost directly"
      ],
      "bestFor": [
        "Private coding and repository evaluation"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "kimi-dev-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "kimi-dev-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "kimi-dev-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "kimi-dev-release"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "kimi-dev-release"
      ]
    },
    {
      "id": "kimi-k2-instruct",
      "providerId": "kimi",
      "name": "Kimi K2 Instruct",
      "family": "Kimi K2",
      "releaseDate": "2025-07-11",
      "status": "active",
      "access": "open-weights",
      "contextWindowTokens": 128000,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Kimi K2 Instruct records open-weight access for text inputs and a published 128,000-token context window"
      ],
      "cautions": [
        "Hosted token price and infrastructure cost are not represented; measure deployment and operating cost directly"
      ],
      "bestFor": [
        "Private instruction-following evaluation with 128,000-token context"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "kimi-k2-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "kimi-k2-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "kimi-k2-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "kimi-k2-release"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1418,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Kimi K2 Preview (0711)"
        }
      ],
      "officialSourceIds": [
        "kimi-k2-release"
      ]
    },
    {
      "id": "gpt-54",
      "providerId": "openai",
      "name": "GPT-5.4",
      "family": "GPT-5.4",
      "releaseDate": "2026-03-05",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1050000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "GPT-5.4 provides a 1,050,000-token context window and configurable reasoning for complex professional work"
      ],
      "cautions": [
        "Prompts above the provider long-context threshold use higher token rates; reproduce cost and quality on the target workload"
      ],
      "bestFor": [
        "Long-context coding and professional-work evaluations"
      ],
      "pricing": {
        "inputPerMillionUsd": 2.5,
        "outputPerMillionUsd": 15,
        "cachedInputPerMillionUsd": 0.25,
        "sourceId": "openai-gpt-5-4-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2.5,
          "tier": "standard",
          "sourceId": "openai-gpt-5-4-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": 15,
          "tier": "standard",
          "sourceId": "openai-gpt-5-4-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.25,
          "tier": "cached",
          "sourceId": "openai-gpt-5-4-model"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "livebench-overall-2026-06-25",
          "benchmarkSnapshotId": "livebench-overall-2026-06-25-snapshot-2026-06-25",
          "score": 77.97,
          "sourceId": "livebench-2026-06-25-results-19e766a5",
          "measuredAt": "2026-06-25",
          "sourceRow": "gpt-5.4-xhigh"
        },
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 93.3,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-5.4-2026-03-05_xhigh"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 78.6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-5.4-2026-03-05_xhigh"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 45.1,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-5.4-2026-03-05_xhigh"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1475,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-5.4 (High)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1401,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-5.4"
        }
      ],
      "officialSourceIds": [
        "openai-gpt-5-4-model"
      ]
    },
    {
      "id": "gpt-55",
      "providerId": "openai",
      "name": "GPT-5.5",
      "family": "GPT-5.5",
      "releaseDate": "2026-04-23",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1050000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "GPT-5.5 provides a 1,050,000-token context window for complex coding and professional work"
      ],
      "cautions": [
        "Prompts above the provider long-context threshold use higher token rates; validate the selected reasoning effort against production cost"
      ],
      "bestFor": [
        "High-capability API evaluations with long context"
      ],
      "pricing": {
        "inputPerMillionUsd": 5,
        "outputPerMillionUsd": 30,
        "cachedInputPerMillionUsd": 0.5,
        "sourceId": "openai-gpt-5-5-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 5,
          "tier": "standard",
          "sourceId": "openai-gpt-5-5-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": 30,
          "tier": "standard",
          "sourceId": "openai-gpt-5-5-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.5,
          "tier": "cached",
          "sourceId": "openai-gpt-5-5-model"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "livebench-overall-2026-06-25",
          "benchmarkSnapshotId": "livebench-overall-2026-06-25-snapshot-2026-06-25",
          "score": 79.91,
          "sourceId": "livebench-2026-06-25-results-19e766a5",
          "measuredAt": "2026-06-25",
          "sourceRow": "gpt-5.5-xhigh"
        },
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 90.7,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-5.5_low"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 85.3,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-5.5_xhigh"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 63,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-5.5_xhigh"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1482,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-5.5 (High)"
        }
      ],
      "officialSourceIds": [
        "openai-gpt-5-5-release",
        "openai-gpt-5-5-model"
      ]
    },
    {
      "id": "gpt-56-sol",
      "providerId": "openai",
      "name": "GPT-5.6 Sol",
      "family": "GPT-5.6",
      "releaseDate": "2026-07-09",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1050000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "GPT-5.6 Sol is OpenAI's GPT-5.6 flagship for complex professional work, with a 1.05M-token context window and 128K output tokens",
        "Six reasoning-effort levels from none to max, and the full Responses API tool set including computer use, hosted shell, apply patch and MCP",
        "Epoch AI measured 93.5% on GPQA Diamond and 89.1% on FrontierMath Tiers 1-3, near the top of this dataset"
      ],
      "cautions": [
        "Not in the pinned LiveBench snapshot; its independent scores here come from Epoch AI's own GPQA Diamond, FrontierMath and SimpleQA Verified runs"
      ],
      "bestFor": [
        "Frontier capability evaluations for complex API workflows",
        "Long-horizon coding and research agents that use computer use or MCP tools"
      ],
      "pricing": {
        "inputPerMillionUsd": 4,
        "outputPerMillionUsd": 20,
        "cachedInputPerMillionUsd": 0.4,
        "sourceId": "openai-api-pricing-2026-09"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 4,
          "tier": "standard",
          "sourceId": "openai-api-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 20,
          "tier": "standard",
          "sourceId": "openai-api-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.4,
          "tier": "cached",
          "sourceId": "openai-api-pricing-2026-09"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 93.5,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-5.6-sol_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 89.1,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-5.6-sol_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 69.7,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-5.6-sol_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1485,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-5.6 Sol (xHigh)"
        }
      ],
      "officialSourceIds": [
        "openai-gpt-5-6-release",
        "openai-gpt-5-6-models",
        "openai-api-pricing-2026-09",
        "openai-gpt-5-6-sol-model"
      ]
    },
    {
      "id": "gpt-56-terra",
      "providerId": "openai",
      "name": "GPT-5.6 Terra",
      "family": "GPT-5.6",
      "releaseDate": "2026-07-09",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1050000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "GPT-5.6 Terra is designed for workloads that balance intelligence and cost: $2 input and $12 output per 1M tokens, half of Sol's input price",
        "Same 1.05M-token context, 128K output, six effort levels and Responses API tool set as Sol",
        "Epoch AI measured 93.3% on GPQA Diamond, within a few tenths of Sol, but 43.2% on SimpleQA Verified against Sol's 69.7%"
      ],
      "cautions": [
        "Not in the pinned LiveBench snapshot; its independent scores here come from Epoch AI's own GPQA Diamond, FrontierMath and SimpleQA Verified runs"
      ],
      "bestFor": [
        "Balanced long-context API evaluations",
        "Reasoning-heavy work that does not depend on broad factual recall"
      ],
      "pricing": {
        "inputPerMillionUsd": 2,
        "outputPerMillionUsd": 12,
        "cachedInputPerMillionUsd": 0.2,
        "sourceId": "openai-gpt-5-6-models"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "standard",
          "sourceId": "openai-gpt-5-6-models"
        },
        {
          "unit": "million-output-tokens",
          "usd": 12,
          "tier": "standard",
          "sourceId": "openai-gpt-5-6-models"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.2,
          "tier": "cached",
          "sourceId": "openai-gpt-5-6-models"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 93.3,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-5.6-terra_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 86,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-5.6-terra_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 43.2,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-5.6-terra_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1467,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-5.6 Terra (xHigh)"
        }
      ],
      "officialSourceIds": [
        "openai-gpt-5-6-release",
        "openai-gpt-5-6-models",
        "openai-gpt-5-6-terra-model"
      ]
    },
    {
      "id": "gpt-56-luna",
      "providerId": "openai",
      "name": "GPT-5.6 Luna",
      "family": "GPT-5.6",
      "releaseDate": "2026-07-09",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1050000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "GPT-5.6 Luna is optimized for cost-sensitive workloads at $0.20 input and $1.20 output per 1M tokens",
        "Keeps the 1.05M-token context, 128K output, six effort levels and full tool set of the larger GPT-5.6 tiers",
        "Epoch AI measured 91.6% on GPQA Diamond and 82.1% on FrontierMath Tiers 1-3, strong for its price"
      ],
      "cautions": [
        "Not in the pinned LiveBench snapshot; its independent scores here come from Epoch AI's own GPQA Diamond, FrontierMath and SimpleQA Verified runs"
      ],
      "bestFor": [
        "High-volume API evaluation with published low token prices",
        "Subagents and batch reasoning where cost per call dominates"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.2,
        "outputPerMillionUsd": 1.2,
        "cachedInputPerMillionUsd": 0.02,
        "sourceId": "openai-gpt-5-6-models"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.2,
          "tier": "standard",
          "sourceId": "openai-gpt-5-6-models"
        },
        {
          "unit": "million-output-tokens",
          "usd": 1.2,
          "tier": "standard",
          "sourceId": "openai-gpt-5-6-models"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.02,
          "tier": "cached",
          "sourceId": "openai-gpt-5-6-models"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 91.6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-5.6-luna_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 82.1,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-5.6-luna_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 41,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-5.6-luna_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1453,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-5.6 Luna (xHigh)"
        }
      ],
      "officialSourceIds": [
        "openai-gpt-5-6-release",
        "openai-gpt-5-6-models",
        "openai-gpt-5-6-luna-model"
      ]
    },
    {
      "id": "claude-opus-48",
      "providerId": "anthropic",
      "name": "Claude Opus 4.8",
      "family": "Claude Opus",
      "releaseDate": "2026-05-28",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Claude Opus 4.8 is positioned for coding, agentic tasks, and professional work with a published 1,000,000-token context window"
      ],
      "cautions": [
        "Maximum output length is not represented because the approved source set does not establish one unambiguously"
      ],
      "bestFor": [
        "Long-running agentic and professional-work evaluations"
      ],
      "pricing": {
        "inputPerMillionUsd": 5,
        "outputPerMillionUsd": 25,
        "cachedInputPerMillionUsd": 0.5,
        "sourceId": "anthropic-pricing-2026-09"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 5,
          "tier": "standard",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 25,
          "tier": "standard",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.5,
          "tier": "cached",
          "sourceId": "anthropic-pricing-2026-09"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "livebench-overall-2026-06-25",
          "benchmarkSnapshotId": "livebench-overall-2026-06-25-snapshot-2026-06-25",
          "score": 78.93,
          "sourceId": "livebench-2026-06-25-results-19e766a5",
          "measuredAt": "2026-06-25",
          "sourceRow": "claude-opus-4-8-xhigh-effort"
        },
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 91,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-opus-4-8_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 80,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-opus-4-8_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 53,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-opus-4-8_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1482,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude Opus 4.8 (High)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1556,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude Opus 4.8 (High)"
        }
      ],
      "officialSourceIds": [
        "anthropic-claude-opus-4-8-release",
        "anthropic-pricing-2026-09"
      ]
    },
    {
      "id": "claude-sonnet-5",
      "providerId": "anthropic",
      "name": "Claude Sonnet 5",
      "family": "Claude Sonnet",
      "releaseDate": "2026-06-30",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Claude Sonnet 5 is positioned for agentic execution, tool use, coding, and knowledge work"
      ],
      "cautions": [
        "The launch price is introductory through 2026-08-31; refresh pricing before evaluating workloads after that date"
      ],
      "bestFor": [
        "Agentic coding evaluations with current introductory API pricing"
      ],
      "pricing": {
        "inputPerMillionUsd": 2,
        "outputPerMillionUsd": 10,
        "cachedInputPerMillionUsd": 0.2,
        "sourceId": "anthropic-pricing-2026-09"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "standard",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 10,
          "tier": "standard",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.2,
          "tier": "cached",
          "sourceId": "anthropic-pricing-2026-09"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "livebench-overall-2026-06-25",
          "benchmarkSnapshotId": "livebench-overall-2026-06-25-snapshot-2026-06-25",
          "score": 74.85,
          "sourceId": "livebench-2026-06-25-results-19e766a5",
          "measuredAt": "2026-06-25",
          "sourceRow": "claude-sonnet-5-xhigh-effort"
        },
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 90.5,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-sonnet-5_xhigh"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 65.6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-sonnet-5_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 33.7,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-sonnet-5_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1462,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude Sonnet 5 (High)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1540,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude Sonnet 5 (High)"
        }
      ],
      "officialSourceIds": [
        "anthropic-claude-sonnet-5-release",
        "anthropic-pricing-2026-09"
      ]
    },
    {
      "id": "claude-fable-5",
      "providerId": "anthropic",
      "name": "Claude Fable 5",
      "family": "Claude Fable",
      "releaseDate": "2026-06-09",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Claude Fable 5 is Anthropic's highest-capability widely released model for demanding reasoning and long-horizon agentic work"
      ],
      "cautions": [
        "Fable requires 30-day data retention, and classifier-triggered requests fall back to Claude Opus 4.8"
      ],
      "bestFor": [
        "Long-horizon coding, reasoning, and knowledge-work evaluations"
      ],
      "pricing": {
        "inputPerMillionUsd": 10,
        "outputPerMillionUsd": 50,
        "cachedInputPerMillionUsd": 1,
        "sourceId": "anthropic-pricing-2026-09"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 10,
          "tier": "standard",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 50,
          "tier": "standard",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 1,
          "tier": "cached",
          "sourceId": "anthropic-pricing-2026-09"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 85.9,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-fable-5_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 87,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-fable-5_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 70.7,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-fable-5_xhigh"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1504,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude Fable 5 (High)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1624,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude Fable 5 (High)"
        }
      ],
      "officialSourceIds": [
        "anthropic-claude-fable-5-release",
        "anthropic-models-overview-current",
        "anthropic-pricing-2026-09"
      ]
    },
    {
      "id": "gemini-35-flash",
      "providerId": "google",
      "name": "Gemini 3.5 Flash",
      "family": "Gemini 3.5",
      "releaseDate": "2026-05-19",
      "status": "legacy",
      "deprecation": {
        "date": "2026-10-08",
        "replacementId": "gemini-36-flash",
        "sourceId": "google-gemini-changelog-2026-10-08"
      },
      "access": "api",
      "contextWindowTokens": 1048576,
      "maxOutputTokens": 65536,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Gemini 3.5 Flash supports text, image, audio, and video inputs with a 1,048,576-token input limit"
      ],
      "cautions": [
        "Google deprecated it on 2026-10-08: Google AI Studio no longer offers it, and API requests to gemini-3.5-flash are routed to gemini-3.6-flash",
        "Exact current token prices are not represented because the approved source set does not establish a single untiered rate"
      ],
      "bestFor": [
        "Long-context multimodal and agentic evaluations"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-gemini-3-5-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-gemini-3-5-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-gemini-3-5-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-gemini-3-5-model"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "livebench-overall-2026-06-25",
          "benchmarkSnapshotId": "livebench-overall-2026-06-25-snapshot-2026-06-25",
          "score": 74.64,
          "sourceId": "livebench-2026-06-25-results-19e766a5",
          "measuredAt": "2026-06-25",
          "sourceRow": "gemini-3.5-flash-high"
        },
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 92.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.5-flash_high"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 62.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.5-flash_high"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 66.2,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.5-flash_high"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1478,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 3.5 Flash (High)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1499,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 3.5 Flash (High)"
        }
      ],
      "officialSourceIds": [
        "google-gemini-3-5-changelog",
        "google-gemini-3-5-model",
        "google-gemini-changelog-2026-10-08"
      ]
    },
    {
      "id": "gemini-36-flash",
      "providerId": "google",
      "name": "Gemini 3.6 Flash",
      "family": "Gemini 3.6",
      "releaseDate": "2026-07-21",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1048576,
      "maxOutputTokens": 65536,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Gemini 3.6 Flash supports multimodal inputs and a 1,048,576-token input limit for agentic and coding workflows",
        "Google says it excels at code generation, agentic execution and spatial reasoning, and lists computer use in preview alongside function calling, code execution and search grounding",
        "Epoch AI measured 94.1% on GPQA Diamond"
      ],
      "cautions": [
        "Not in the pinned LiveBench snapshot; its independent scores here come from Epoch AI's own GPQA Diamond, FrontierMath and SimpleQA Verified runs"
      ],
      "bestFor": [
        "Current multimodal agentic evaluations balancing speed and capability"
      ],
      "pricing": {
        "inputPerMillionUsd": 1.5,
        "outputPerMillionUsd": 7.5,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-gemini-3-6-latest"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 1.5,
          "tier": "standard",
          "sourceId": "google-gemini-3-6-latest"
        },
        {
          "unit": "million-output-tokens",
          "usd": 7.5,
          "tier": "standard",
          "sourceId": "google-gemini-3-6-latest"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-gemini-3-6-latest"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 94.1,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.6-flash_high"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 58.9,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.6-flash_high"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 66.2,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.6-flash_high"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1483,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 3.6 Flash (High)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1538,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 3.6 Flash (High)"
        }
      ],
      "officialSourceIds": [
        "google-gemini-3-6-model",
        "google-gemini-3-6-latest"
      ]
    },
    {
      "id": "gemini-35-flash-lite",
      "providerId": "google",
      "name": "Gemini 3.5 Flash-Lite",
      "family": "Gemini 3.5",
      "releaseDate": "2026-07-21",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1048576,
      "maxOutputTokens": 65536,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Gemini 3.5 Flash-Lite is optimized for high-throughput, low-cost execution of subagent tasks and document parsing",
        "Reads text, images, video, audio and PDF in a 1,048,576-token window with 65,536 output tokens, and supports thinking, function calling, code execution, search grounding and caching"
      ],
      "cautions": [
        "The represented prices are standard API rates; batch, flex, and priority consumption use different rates"
      ],
      "bestFor": [
        "High-volume extraction, structured parsing, and cost-sensitive subagent trials"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.3,
        "outputPerMillionUsd": 2.5,
        "cachedInputPerMillionUsd": 0.03,
        "sourceId": "google-gemini-3-5-flash-lite-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.3,
          "tier": "standard",
          "sourceId": "google-gemini-3-5-flash-lite-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 2.5,
          "tier": "standard",
          "sourceId": "google-gemini-3-5-flash-lite-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.03,
          "tier": "cached",
          "sourceId": "google-gemini-3-5-flash-lite-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 83.3,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.5-flash-lite_high"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 26,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.5-flash-lite_high"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1456,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 3.5 Flash-Lite"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1434,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 3.5 Flash-Lite"
        }
      ],
      "officialSourceIds": [
        "google-gemini-3-5-flash-lite-model",
        "google-gemini-3-5-flash-lite-pricing",
        "google-gemini-3-5-changelog"
      ]
    },
    {
      "id": "grok-420",
      "providerId": "xai",
      "name": "Grok 4.20",
      "family": "Grok 4",
      "releaseDate": "2026-03-10",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Grok 4.20 supports text and image input and is available in reasoning, non-reasoning, and multi-agent API variants"
      ],
      "cautions": [
        "The published base rates double when requests reach the provider's 200,000-token long-context threshold"
      ],
      "bestFor": [
        "Comparing single-agent and multi-agent reasoning on long-context workloads"
      ],
      "pricing": {
        "inputPerMillionUsd": 1.25,
        "outputPerMillionUsd": 2.5,
        "cachedInputPerMillionUsd": 0.2,
        "sourceId": "xai-current-model-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 1.25,
          "tier": "standard",
          "sourceId": "xai-current-model-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 2.5,
          "tier": "standard",
          "sourceId": "xai-current-model-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.2,
          "tier": "cached",
          "sourceId": "xai-current-model-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 89.3,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "grok-4.20-0309-reasoning"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 44.9,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "grok-4.20-0309-reasoning"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 30.2,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "grok-4.20-0309-reasoning"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1472,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Grok 4.20 Beta (Reasoning, 0309)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1375,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Grok 4.20 Beta (Reasoning, 0309)"
        }
      ],
      "officialSourceIds": [
        "xai-release-notes-current",
        "xai-grok-4-20-system-card",
        "xai-current-model-pricing"
      ]
    },
    {
      "id": "grok-43",
      "providerId": "xai",
      "name": "Grok 4.3",
      "family": "Grok 4",
      "releaseDate": null,
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Grok 4.3 is xAI's general-purpose default with configurable reasoning and strong tool calling"
      ],
      "cautions": [
        "The published base rates double when requests reach the provider's 200,000-token long-context threshold"
      ],
      "bestFor": [
        "General-purpose reasoning, instruction following, and agentic tool-use evaluations"
      ],
      "pricing": {
        "inputPerMillionUsd": 1.25,
        "outputPerMillionUsd": 2.5,
        "cachedInputPerMillionUsd": 0.2,
        "sourceId": "xai-grok-4-3-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 1.25,
          "tier": "standard",
          "sourceId": "xai-grok-4-3-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": 2.5,
          "tier": "standard",
          "sourceId": "xai-grok-4-3-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.2,
          "tier": "cached",
          "sourceId": "xai-grok-4-3-model"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 88.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "grok-4.3_high"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 42.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "grok-4.3_high"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 33.2,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "grok-4.3_high"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1442,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Grok 4.3"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1358,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Grok 4.3"
        }
      ],
      "officialSourceIds": [
        "xai-grok-4-3-model",
        "xai-grok-4-3-public-availability",
        "xai-inference-models-api",
        "xai-may-15-retirement"
      ]
    },
    {
      "id": "grok-45",
      "providerId": "xai",
      "name": "Grok 4.5",
      "family": "Grok 4",
      "releaseDate": "2026-07-16",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 500000,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Grok 4.5 is positioned for coding, agentic tasks, and knowledge work with a 500,000-token context window",
        "xAI says it is served at about 80 tokens per second and solves tasks in under half the steps, at $2 input and $6 output per 1M tokens"
      ],
      "cautions": [
        "The published base rates increase for requests at or above the provider long-context threshold"
      ],
      "bestFor": [
        "Coding and agentic API evaluations below the long-context pricing threshold"
      ],
      "pricing": {
        "inputPerMillionUsd": 2,
        "outputPerMillionUsd": 6,
        "cachedInputPerMillionUsd": 0.3,
        "sourceId": "xai-grok-4-5-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "standard",
          "sourceId": "xai-grok-4-5-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": 6,
          "tier": "standard",
          "sourceId": "xai-grok-4-5-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.3,
          "tier": "cached",
          "sourceId": "xai-grok-4-5-model"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 93.4,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "grok-4.5_high"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 57.2,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "grok-4.5_high"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 48.3,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "grok-4.5_high"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1465,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Grok 4.5"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1553,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Grok 4.5"
        }
      ],
      "officialSourceIds": [
        "xai-grok-4-5-release",
        "xai-grok-4-5-model"
      ]
    },
    {
      "id": "grok-46",
      "providerId": "xai",
      "name": "Grok 4.6",
      "family": "Grok 4",
      "releaseDate": "2026-08-12",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 500000,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Grok 4.6 is positioned for coding, agentic tasks, and knowledge work; its API documentation also lists function calling and structured outputs",
        "xAI says it focuses on long-running agents that stay with complex tasks across many steps, with more self-testing and verification than Grok 4.5",
        "Epoch AI measured 94.0% on GPQA Diamond and 66.0% on FrontierMath Tiers 1-3"
      ],
      "cautions": [
        "Published token rates increase for all tokens in a request once its prompt reaches xAI's 200,000-token long-context threshold"
      ],
      "bestFor": [
        "Coding, agentic, and knowledge-work API evaluations with controlled long-context cost"
      ],
      "pricing": {
        "inputPerMillionUsd": 2,
        "outputPerMillionUsd": 6,
        "cachedInputPerMillionUsd": 0.5,
        "sourceId": "xai-grok-4-6-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "standard",
          "sourceId": "xai-grok-4-6-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": 6,
          "tier": "standard",
          "sourceId": "xai-grok-4-6-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.5,
          "tier": "cached",
          "sourceId": "xai-grok-4-6-model"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 94,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "grok-4.6_high"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 66,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "grok-4.6_xhigh"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 49.3,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "grok-4.6_high"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1455,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Grok 4.6 (High)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1616,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Grok 4.6 (High)"
        }
      ],
      "officialSourceIds": [
        "xai-grok-4-6-release",
        "xai-grok-4-6-model",
        "xai-current-model-pricing"
      ]
    },
    {
      "id": "gpt-image-2",
      "providerId": "openai",
      "name": "GPT Image 2",
      "family": "GPT Image",
      "releaseDate": null,
      "status": "active",
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "image"
      ],
      "capabilities": [
        "image-generation",
        "image-editing",
        "image-to-image"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "image"
      ],
      "strengths": [
        "Official documentation supports image generation, editing, and high-fidelity image inputs"
      ],
      "cautions": [
        "Arena rates the newer GPT Image 2.5 Sunburst and Flare higher for both generation and editing"
      ],
      "bestFor": [
        "API image generation and editing"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "specialist-openai-gpt-image-2"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-openai-gpt-image-2"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-openai-gpt-image-2"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "specialist-openai-gpt-image-2"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-image",
          "benchmarkSnapshotId": "arena-text-to-image-snapshot-2026-11-04",
          "score": 1382,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "GPT-Image-2 (Medium)"
        },
        {
          "benchmarkId": "arena-image-edit",
          "benchmarkSnapshotId": "arena-image-edit-snapshot-2026-11-04",
          "score": 1460,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "GPT-Image-2 (Medium)"
        }
      ],
      "officialSourceIds": [
        "specialist-openai-gpt-image-2"
      ]
    },
    {
      "id": "gemini-3.1-flash-image",
      "providerId": "google",
      "name": "Gemini 3.1 Flash Image",
      "family": "Gemini Image",
      "releaseDate": "2026-05-28",
      "status": "legacy",
      "deprecation": {
        "date": "2026-10-06",
        "replacementId": "gemini-nano-banana-2-1",
        "sourceId": "google-gemini-changelog-2026-10-06"
      },
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "image",
        "multimodal"
      ],
      "capabilities": [
        "image-generation",
        "image-editing",
        "image-to-image"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text",
        "image"
      ],
      "strengths": [
        "Google documents native text-and-image generation and editing with up to 4K output"
      ],
      "cautions": [
        "Google deprecated it on 2026-10-06 with no shutdown date announced and recommends migrating to Nano Banana 2.1 (gemini-nano-banana-2.1)"
      ],
      "bestFor": [
        "Interleaved text and image generation"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.5,
        "outputPerMillionUsd": 3,
        "cachedInputPerMillionUsd": null,
        "sourceId": "specialist-google-gemini-3-1-flash-image-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.5,
          "tier": "standard",
          "sourceId": "specialist-google-gemini-3-1-flash-image-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 3,
          "tier": "standard",
          "sourceId": "specialist-google-gemini-3-1-flash-image-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "specialist-google-gemini-3-1-flash-image-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 60,
          "tier": "standard-image-output",
          "sourceId": "specialist-google-gemini-3-1-flash-image-pricing"
        },
        {
          "unit": "image",
          "usd": 0.045,
          "tier": "standard-0.5k",
          "sourceId": "specialist-google-gemini-3-1-flash-image-pricing"
        },
        {
          "unit": "image",
          "usd": 0.067,
          "tier": "standard-1k",
          "sourceId": "specialist-google-gemini-3-1-flash-image-pricing"
        },
        {
          "unit": "image",
          "usd": 0.101,
          "tier": "standard-2k",
          "sourceId": "specialist-google-gemini-3-1-flash-image-pricing"
        },
        {
          "unit": "image",
          "usd": 0.151,
          "tier": "standard-4k",
          "sourceId": "specialist-google-gemini-3-1-flash-image-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-image",
          "benchmarkSnapshotId": "arena-text-to-image-snapshot-2026-11-04",
          "score": 1257,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "Gemini 3.1 Flash Image (nano-banana-2, Web Search)"
        },
        {
          "benchmarkId": "arena-image-edit",
          "benchmarkSnapshotId": "arena-image-edit-snapshot-2026-11-04",
          "score": 1387,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "Gemini 3.1 Flash Image (nano-banana-2, Web Search)"
        }
      ],
      "officialSourceIds": [
        "specialist-google-native-image-capabilities",
        "specialist-google-gemini-3-1-flash-image-pricing",
        "google-gemini-deprecations",
        "google-gemini-changelog-2026-10-06"
      ]
    },
    {
      "id": "grok-imagine-image",
      "providerId": "xai",
      "name": "grok-imagine-image",
      "family": "Grok Imagine",
      "releaseDate": null,
      "status": "active",
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "image"
      ],
      "capabilities": [
        "image-generation",
        "image-editing",
        "image-to-image"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "image"
      ],
      "strengths": [
        "xAI documents text and image inputs for image generation and editing"
      ],
      "cautions": [
        "Arena rates the newer Grok Imagine Image 2.0 well above it for both generation and editing"
      ],
      "bestFor": [
        "Grok Imagine image workflows"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "specialist-xai-imagine-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-xai-imagine-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-xai-imagine-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "specialist-xai-imagine-pricing"
        },
        {
          "unit": "image",
          "usd": 0.02,
          "tier": "output-1k",
          "sourceId": "specialist-xai-imagine-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-image",
          "benchmarkSnapshotId": "arena-text-to-image-snapshot-2026-11-04",
          "score": 1169,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "Grok Imagine Image"
        },
        {
          "benchmarkId": "arena-image-edit",
          "benchmarkSnapshotId": "arena-image-edit-snapshot-2026-11-04",
          "score": 1328,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "Grok Imagine Image"
        }
      ],
      "officialSourceIds": [
        "specialist-xai-imagine-image",
        "specialist-xai-imagine-pricing"
      ]
    },
    {
      "id": "stable-image-ultra",
      "providerId": "stability-ai",
      "name": "Stable Image Ultra",
      "family": "Stable Image",
      "releaseDate": null,
      "status": "active",
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "image"
      ],
      "capabilities": [
        "image-generation"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "image"
      ],
      "strengths": [
        "Stability documents its flagship image service based on Stable Diffusion 3.5 Large"
      ],
      "cautions": [
        "No exact compatible public benchmark snapshot is recorded"
      ],
      "bestFor": [
        "Hosted high-detail image generation"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "specialist-stability-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-stability-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-stability-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "specialist-stability-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "specialist-stability-pricing"
      ]
    },
    {
      "id": "flux-pro-1.1",
      "providerId": "black-forest-labs",
      "name": "FLUX1.1 [pro]",
      "family": "FLUX.1",
      "releaseDate": null,
      "status": "active",
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "image"
      ],
      "capabilities": [
        "image-generation"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "image"
      ],
      "strengths": [
        "Black Forest Labs documents the flux-pro-1.1 text-to-image endpoint"
      ],
      "cautions": [
        "An older model: Arena rates the FLUX.2 models from Black Forest Labs well above it"
      ],
      "bestFor": [
        "Hosted FLUX text-to-image generation"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "specialist-bfl-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-bfl-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-bfl-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "specialist-bfl-pricing"
        },
        {
          "unit": "image",
          "usd": 0.04,
          "tier": "standard",
          "sourceId": "specialist-bfl-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-image",
          "benchmarkSnapshotId": "arena-text-to-image-snapshot-2026-11-04",
          "score": 1016,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "FLUX1.1 [pro]"
        }
      ],
      "officialSourceIds": [
        "specialist-bfl-models",
        "specialist-bfl-pricing"
      ]
    },
    {
      "id": "veo-3.1-generate-preview",
      "providerId": "google",
      "name": "Veo 3.1 Preview",
      "family": "Veo",
      "releaseDate": "2025-10-15",
      "status": "active",
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "video",
        "audio"
      ],
      "taskFamilies": [
        "video"
      ],
      "capabilities": [
        "text-to-video",
        "image-to-video",
        "sound-generation"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "video",
        "audio"
      ],
      "strengths": [
        "Google lists the exact Veo 3.1 preview identifier for video generation with audio"
      ],
      "cautions": [
        "Preview models may change and have stricter rate limits"
      ],
      "bestFor": [
        "Gemini API video generation with audio"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "specialist-google-native-media"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-google-native-media"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-google-native-media"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "specialist-google-native-media"
        },
        {
          "unit": "video-second",
          "usd": 0.4,
          "tier": "720p-or-1080p-with-audio",
          "sourceId": "specialist-google-native-media"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-video",
          "benchmarkSnapshotId": "arena-text-to-video-snapshot-2026-11-05",
          "score": 1363,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-05",
          "sourceRow": "Veo 3.1 (Audio)"
        },
        {
          "benchmarkId": "arena-image-to-video",
          "benchmarkSnapshotId": "arena-image-to-video-snapshot-2026-11-05",
          "score": 1399,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-05",
          "sourceRow": "Veo 3.1 (Audio)"
        }
      ],
      "officialSourceIds": [
        "specialist-google-native-media",
        "google-gemini-deprecations"
      ]
    },
    {
      "id": "gen4.5",
      "providerId": "runway",
      "name": "Gen-4.5",
      "family": "Runway Gen",
      "releaseDate": null,
      "status": "active",
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "video"
      ],
      "taskFamilies": [
        "video"
      ],
      "capabilities": [
        "text-to-video",
        "image-to-video"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "video"
      ],
      "strengths": [
        "Runway documents the exact gen4.5 identifier with text or image input and video output"
      ],
      "cautions": [
        "No per-second price is recorded here, so cost comparisons leave it out"
      ],
      "bestFor": [
        "Runway API video generation"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "specialist-runway-models"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-runway-models"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-runway-models"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "specialist-runway-models"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-video",
          "benchmarkSnapshotId": "arena-text-to-video-snapshot-2026-11-05",
          "score": 1223,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-05",
          "sourceRow": "Runway Gen-4.5"
        }
      ],
      "officialSourceIds": [
        "specialist-runway-models"
      ]
    },
    {
      "id": "grok-imagine-video",
      "providerId": "xai",
      "name": "grok-imagine-video",
      "family": "Grok Imagine",
      "releaseDate": null,
      "status": "active",
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "video"
      ],
      "taskFamilies": [
        "video"
      ],
      "capabilities": [
        "text-to-video",
        "image-to-video",
        "video-editing"
      ],
      "inputModalities": [
        "text",
        "image",
        "video"
      ],
      "outputModalities": [
        "video"
      ],
      "strengths": [
        "xAI documents text, image, and video inputs for video generation and editing"
      ],
      "cautions": [
        "The newer Grok Imagine Video 1.5 and 1.5 Lite generate video; this model remains the one for editing and extending clips"
      ],
      "bestFor": [
        "Grok Imagine video generation and editing"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "specialist-xai-imagine-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-xai-imagine-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-xai-imagine-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "specialist-xai-imagine-pricing"
        },
        {
          "unit": "video-second",
          "usd": 0.05,
          "tier": "480p",
          "sourceId": "specialist-xai-imagine-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-video",
          "benchmarkSnapshotId": "arena-text-to-video-snapshot-2026-11-05",
          "score": 1341,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-05",
          "sourceRow": "Grok Imagine Video (720p)"
        },
        {
          "benchmarkId": "arena-image-to-video",
          "benchmarkSnapshotId": "arena-image-to-video-snapshot-2026-11-05",
          "score": 1412,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-05",
          "sourceRow": "Grok Imagine Video (720p)"
        }
      ],
      "officialSourceIds": [
        "specialist-xai-imagine-video",
        "specialist-xai-imagine-pricing"
      ]
    },
    {
      "id": "gpt-realtime-1.5",
      "providerId": "openai",
      "name": "GPT-Realtime-1.5",
      "family": "GPT Realtime",
      "releaseDate": null,
      "status": "active",
      "access": "api",
      "contextWindowTokens": 32000,
      "maxOutputTokens": 4096,
      "modalities": [
        "text",
        "image",
        "audio"
      ],
      "taskFamilies": [
        "audio",
        "multimodal"
      ],
      "capabilities": [
        "audio-understanding",
        "realtime-voice",
        "speech-to-text",
        "text-to-speech"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio"
      ],
      "outputModalities": [
        "text",
        "audio"
      ],
      "strengths": [
        "OpenAI documents realtime text and audio input and output for voice agents"
      ],
      "cautions": [
        "OpenAI released GPT-Realtime-2 on 2026-05-07 with configurable reasoning; no independent voice-quality score is recorded for either"
      ],
      "bestFor": [
        "Realtime voice agents"
      ],
      "pricing": {
        "inputPerMillionUsd": 4,
        "outputPerMillionUsd": 16,
        "cachedInputPerMillionUsd": 0.4,
        "sourceId": "specialist-openai-realtime-1-5"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 4,
          "tier": "standard",
          "sourceId": "specialist-openai-realtime-1-5"
        },
        {
          "unit": "million-output-tokens",
          "usd": 16,
          "tier": "standard",
          "sourceId": "specialist-openai-realtime-1-5"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.4,
          "tier": "cached",
          "sourceId": "specialist-openai-realtime-1-5"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "specialist-openai-realtime-1-5"
      ]
    },
    {
      "id": "gpt-audio-1.5",
      "providerId": "openai",
      "name": "gpt-audio-1.5",
      "family": "GPT Audio",
      "releaseDate": null,
      "status": "active",
      "access": "api",
      "contextWindowTokens": 128000,
      "maxOutputTokens": 16384,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio",
        "multimodal"
      ],
      "capabilities": [
        "audio-understanding",
        "speech-to-text",
        "text-to-speech"
      ],
      "inputModalities": [
        "text",
        "audio"
      ],
      "outputModalities": [
        "text",
        "audio"
      ],
      "strengths": [
        "OpenAI documents audio and text inputs and outputs through Chat Completions"
      ],
      "cautions": [
        "No exact compatible public benchmark snapshot is recorded"
      ],
      "bestFor": [
        "Audio-in and audio-out conversations"
      ],
      "pricing": {
        "inputPerMillionUsd": 2.5,
        "outputPerMillionUsd": 10,
        "cachedInputPerMillionUsd": null,
        "sourceId": "specialist-openai-audio-1-5"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2.5,
          "tier": "standard",
          "sourceId": "specialist-openai-audio-1-5"
        },
        {
          "unit": "million-output-tokens",
          "usd": 10,
          "tier": "standard",
          "sourceId": "specialist-openai-audio-1-5"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "specialist-openai-audio-1-5"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "specialist-openai-audio-1-5"
      ]
    },
    {
      "id": "gpt-4o-transcribe",
      "providerId": "openai",
      "name": "GPT-4o Transcribe",
      "family": "GPT Transcribe",
      "releaseDate": "2025-03-20",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 16000,
      "maxOutputTokens": 2000,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "speech-to-text"
      ],
      "inputModalities": [
        "text",
        "audio"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "OpenAI documents speech-to-text with improved word error rate over Whisper"
      ],
      "cautions": [
        "OpenAI announced its deprecation on 2026-08-26 and recommends GPT Transcribe, which costs $0.0045 per minute against about $0.006 here"
      ],
      "bestFor": [
        "Speech transcription"
      ],
      "pricing": {
        "inputPerMillionUsd": 2.5,
        "outputPerMillionUsd": 10,
        "cachedInputPerMillionUsd": null,
        "sourceId": "specialist-openai-transcribe"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2.5,
          "tier": "standard",
          "sourceId": "specialist-openai-transcribe"
        },
        {
          "unit": "million-output-tokens",
          "usd": 10,
          "tier": "standard",
          "sourceId": "specialist-openai-transcribe"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "specialist-openai-transcribe"
        },
        {
          "unit": "audio-minute",
          "usd": 0.006,
          "tier": "estimated",
          "sourceId": "specialist-openai-transcribe"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "specialist-openai-transcribe",
        "openai-api-changelog"
      ]
    },
    {
      "id": "tts-1",
      "providerId": "openai",
      "name": "TTS-1",
      "family": "OpenAI TTS",
      "releaseDate": null,
      "status": "active",
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "text-to-speech"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "audio"
      ],
      "strengths": [
        "OpenAI documents TTS-1 for realtime text-to-speech use cases"
      ],
      "cautions": [
        "OpenAI no longer lists TTS-1 on its models overview, though its model page and $15 per 1M characters price remain"
      ],
      "bestFor": [
        "Low-latency text-to-speech"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "specialist-openai-tts-1"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-openai-tts-1"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-openai-tts-1"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "specialist-openai-tts-1"
        },
        {
          "unit": "character",
          "usd": 0.000015,
          "tier": "standard",
          "sourceId": "specialist-openai-tts-1"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "specialist-openai-tts-1"
      ]
    },
    {
      "id": "eleven_v3",
      "providerId": "elevenlabs",
      "name": "Eleven v3",
      "family": "Eleven TTS",
      "releaseDate": null,
      "status": "active",
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "text-to-speech"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "audio"
      ],
      "strengths": [
        "ElevenLabs documents expressive speech synthesis across more than 70 languages"
      ],
      "cautions": [
        "ElevenLabs now calls it a previous-generation model; Eleven v4 replaces it at the same $0.08 per 1,000 characters"
      ],
      "bestFor": [
        "Expressive multilingual speech"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "specialist-elevenlabs-models"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-elevenlabs-models"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-elevenlabs-models"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "specialist-elevenlabs-models"
        },
        {
          "unit": "character",
          "usd": 0.00008,
          "tier": "standard",
          "sourceId": "elevenlabs-api-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "specialist-elevenlabs-models",
        "elevenlabs-api-pricing"
      ]
    },
    {
      "id": "eleven_multilingual_v2",
      "providerId": "elevenlabs",
      "name": "Eleven Multilingual v2",
      "family": "Eleven TTS",
      "releaseDate": null,
      "status": "active",
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "text-to-speech"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "audio"
      ],
      "strengths": [
        "ElevenLabs documents stable, lifelike multilingual long-form speech"
      ],
      "cautions": [
        "ElevenLabs now calls it a previous-generation model, priced the same as Eleven v4"
      ],
      "bestFor": [
        "Long-form multilingual narration"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "specialist-elevenlabs-models"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-elevenlabs-models"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-elevenlabs-models"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "specialist-elevenlabs-models"
        },
        {
          "unit": "character",
          "usd": 0.00008,
          "tier": "standard",
          "sourceId": "elevenlabs-api-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "specialist-elevenlabs-models",
        "elevenlabs-api-pricing"
      ]
    },
    {
      "id": "eleven_flash_v2_5",
      "providerId": "elevenlabs",
      "name": "Eleven Flash v2.5",
      "family": "Eleven TTS",
      "releaseDate": null,
      "status": "active",
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "text-to-speech",
        "realtime-voice"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "audio"
      ],
      "strengths": [
        "ElevenLabs documents approximately 75 ms model latency across 32 languages"
      ],
      "cautions": [
        "Built for latency rather than expressiveness; Eleven v4 Turbo now sells at the same $0.04 per 1,000 characters"
      ],
      "bestFor": [
        "Low-latency multilingual speech"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "specialist-elevenlabs-models"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-elevenlabs-models"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-elevenlabs-models"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "specialist-elevenlabs-models"
        },
        {
          "unit": "character",
          "usd": 0.00004,
          "tier": "standard",
          "sourceId": "elevenlabs-api-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "specialist-elevenlabs-models",
        "elevenlabs-api-pricing"
      ]
    },
    {
      "id": "scribe_v2",
      "providerId": "elevenlabs",
      "name": "Scribe v2",
      "family": "Scribe",
      "releaseDate": "2026-01-09",
      "status": "active",
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "audio",
        "text"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "speech-to-text",
        "diarization",
        "audio-understanding"
      ],
      "inputModalities": [
        "audio"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "ElevenLabs documents transcription in more than 90 languages with diarization"
      ],
      "cautions": [
        "$0.22 per hour of audio; keyterm prompting and transcript editing add $0.05 and $0.07 per hour"
      ],
      "bestFor": [
        "Batch transcription and diarization"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "specialist-elevenlabs-models"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-elevenlabs-models"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-elevenlabs-models"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "specialist-elevenlabs-models"
        },
        {
          "unit": "audio-minute",
          "usd": 0.003667,
          "tier": "standard",
          "sourceId": "elevenlabs-api-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "specialist-elevenlabs-models",
        "specialist-elevenlabs-scribe-v2-release",
        "elevenlabs-api-pricing"
      ]
    },
    {
      "id": "scribe_v2_realtime",
      "providerId": "elevenlabs",
      "name": "Scribe v2 Realtime",
      "family": "Scribe",
      "releaseDate": "2025-11-11",
      "status": "active",
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "audio",
        "text"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "speech-to-text",
        "realtime-voice",
        "audio-understanding"
      ],
      "inputModalities": [
        "audio"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "ElevenLabs documents realtime transcription in more than 90 languages"
      ],
      "cautions": [
        "$0.39 per hour of audio, about 1.8 times batch Scribe v2"
      ],
      "bestFor": [
        "Live transcription"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "specialist-elevenlabs-models"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-elevenlabs-models"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "specialist-elevenlabs-models"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "specialist-elevenlabs-models"
        },
        {
          "unit": "audio-minute",
          "usd": 0.0065,
          "tier": "standard",
          "sourceId": "elevenlabs-api-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "specialist-elevenlabs-models",
        "specialist-elevenlabs-scribe-realtime-release",
        "elevenlabs-api-pricing"
      ]
    },
    {
      "id": "gpt-6-astra",
      "providerId": "openai",
      "name": "GPT-6 Astra",
      "family": "GPT-6",
      "releaseDate": "2026-09-03",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1050000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "OpenAI positions GPT-6 Astra as its frontier model for reasoning, agentic work, and computer use, with a 1,050,000-token context window and 128,000-token maximum output",
        "OpenAI reports the strongest published deep-context retention of any frontier model on MRCR v2 (8 needles): 100% up to 512K tokens and 96.3% in the 512K-1M band",
        "OpenAI reports saturation-level scores on FrontierMath Tier 4 (97.6%) and, under its own provider adapter harness, ARC-AGI-3 (99.9%)"
      ],
      "cautions": [
        "Long-context billing is a cliff, not a taper: a request over the long-context threshold is rebilled in full at $20 per million input and $75 per million output, so a prompt that crosses it costs roughly double",
        "GPT-6 Astra is the first publicly released model OpenAI classifies as crossing the Critical cybersecurity threshold under its own Preparedness Framework, which carries additional deployment safeguards",
        "Vendor-reported benchmark figures come from OpenAI’s own launch material and its own harness; they are not independently reproduced here"
      ],
      "bestFor": [
        "Frontier reasoning and agentic evaluations where deep-context retrieval accuracy is the binding constraint",
        "Workloads that can be kept under the long-context threshold, where the effective rate stays at $10 / $50"
      ],
      "pricing": {
        "inputPerMillionUsd": 10,
        "outputPerMillionUsd": 50,
        "cachedInputPerMillionUsd": 1,
        "sourceId": "openai-api-pricing-2026-09"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 10,
          "tier": "standard",
          "sourceId": "openai-api-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 50,
          "tier": "standard",
          "sourceId": "openai-api-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 1,
          "tier": "cached",
          "sourceId": "openai-api-pricing-2026-09"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 95.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-6-astra_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 93.7,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-6-astra_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 75.6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-6-astra_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1476,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-6 Astra (Max)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1773,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-6 Astra (Max)"
        }
      ],
      "officialSourceIds": [
        "openai-gpt-6-astra-model",
        "openai-api-pricing-2026-09"
      ]
    },
    {
      "id": "claude-fable-5-1",
      "providerId": "anthropic",
      "name": "Claude Fable 5.1",
      "family": "Claude Fable",
      "releaseDate": "2026-09-01",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Anthropic’s highest-capability generally available model, for demanding reasoning and long-horizon agentic work, with a 1M-token context window and 128,000-token maximum output",
        "Cache reads are priced at 0.025x base input ($0.25 per million) rather than the 0.1x every other Claude model uses, which is the single largest cost lever for agentic loops that replay a large prompt",
        "Anthropic reports 52.6% on Terminal-Bench-Science 0.1 against 24.7% for Fable 5 and 29.0% for Opus 5",
        "Adaptive thinking is always on and the API effort default is `high`"
      ],
      "cautions": [
        "Base input and output pricing is unchanged from Fable 5 at $10 / $50; the saving is entirely in cache reads, so a workload with little cache reuse sees no discount",
        "The 4.7-generation tokenizer produces roughly 30% more tokens for the same text than Claude Sonnet 4.6 and earlier, so per-token rates are not directly comparable across generations",
        "Anthropic’s launch comparisons are self-reported and were not reproduced independently here"
      ],
      "bestFor": [
        "Long-horizon agentic coding where a large prompt prefix is replayed across many turns",
        "Work where Claude Opus 5 at higher effort has already been evaluated and fell short"
      ],
      "pricing": {
        "inputPerMillionUsd": 10,
        "outputPerMillionUsd": 50,
        "cachedInputPerMillionUsd": 0.25,
        "sourceId": "anthropic-pricing-2026-09"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 10,
          "tier": "standard",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 50,
          "tier": "standard",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.25,
          "tier": "cached",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 12.5,
          "tier": "cache-write-5m",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 20,
          "tier": "cache-write-1h",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 5,
          "tier": "batch",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 25,
          "tier": "batch",
          "sourceId": "anthropic-pricing-2026-09"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 90.2,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-fable-5-1_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 70.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-fable-5-1_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1504,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude Fable 5.1 (Max)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1746,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude Fable 5.1 (Max)"
        }
      ],
      "officialSourceIds": [
        "anthropic-models-overview-2026-09",
        "anthropic-pricing-2026-09"
      ]
    },
    {
      "id": "claude-mythos-5-1",
      "providerId": "anthropic",
      "name": "Claude Mythos 5.1",
      "family": "Claude Mythos",
      "releaseDate": "2026-09-01",
      "status": "active",
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Shipped alongside Claude Fable 5.1 and listed at the same rates: $10 per million input, $50 per million output, and the same 0.025x cache-read multiplier ($0.25 per million)",
        "Carries the same reduced cache-read pricing as Fable 5.1, which no other Claude model has"
      ],
      "cautions": [
        "Limited availability: access is restricted to organisations vetted through Anthropic’s Glasswing programme, so it cannot be assumed available for a general evaluation",
        "Anthropic’s public model comparison table does not list a context window or maximum output for Mythos 5.1, so both are recorded as unknown here rather than inferred from Fable 5.1"
      ],
      "bestFor": [
        "Organisations already admitted to Anthropic’s limited-availability programme that need frontier capability at Fable 5.1 rates"
      ],
      "pricing": {
        "inputPerMillionUsd": 10,
        "outputPerMillionUsd": 50,
        "cachedInputPerMillionUsd": 0.25,
        "sourceId": "anthropic-pricing-2026-09"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 10,
          "tier": "standard",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 50,
          "tier": "standard",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.25,
          "tier": "cached",
          "sourceId": "anthropic-pricing-2026-09"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "anthropic-pricing-2026-09",
        "anthropic-glasswing-2026-09"
      ]
    },
    {
      "id": "claude-opus-5",
      "providerId": "anthropic",
      "name": "Claude Opus 5",
      "family": "Claude Opus",
      "releaseDate": "2026-07-24",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Anthropic’s recommended default for most workloads, at half Fable 5.1’s token price ($5 / $25) with the same 1M-token context window and 128,000-token maximum output",
        "Exposes five effort settings (low, medium, high, xhigh, max), which trade capability against token consumption without changing model",
        "One of only two models with a Fast mode research preview, billed at $10 / $50 across the full context window"
      ],
      "cautions": [
        "Fast mode pricing applies to the whole request, including above 200K input tokens, so it doubles the effective rate rather than adding a surcharge",
        "Cache reads are $0.50 per million (the standard 0.1x), five times Fable 5.1’s $0.25, even though Opus 5 is the cheaper model on base tokens. A cache-heavy agent loop can invert the expected cost ordering between the two."
      ],
      "bestFor": [
        "Complex agentic coding and enterprise work as the general-purpose default before reaching for Fable 5.1",
        "Latency-sensitive work that can justify Fast mode pricing"
      ],
      "pricing": {
        "inputPerMillionUsd": 5,
        "outputPerMillionUsd": 25,
        "cachedInputPerMillionUsd": 0.5,
        "sourceId": "anthropic-pricing-2026-09"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 5,
          "tier": "standard",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 25,
          "tier": "standard",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.5,
          "tier": "cached",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 2.5,
          "tier": "batch",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 12.5,
          "tier": "batch",
          "sourceId": "anthropic-pricing-2026-09"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 93.9,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-opus-5_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 85.6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-opus-5_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 59.9,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-opus-5_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1492,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude Opus 5 (High)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1691,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude Opus 5 (Max)"
        }
      ],
      "officialSourceIds": [
        "anthropic-models-overview-2026-09",
        "anthropic-pricing-2026-09"
      ]
    },
    {
      "id": "gemini-38-flash",
      "providerId": "google",
      "name": "Gemini 3.8 Flash",
      "family": "Gemini 3.8",
      "releaseDate": "2026-09-02",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1048576,
      "maxOutputTokens": 65536,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "The cheapest frontier-tier option in this dataset by a wide margin: $0.75 per million input and $3.75 per million output, roughly one thirteenth of GPT-6 Astra on both",
        "Accepts text, image, audio, video and PDF input with a 1,048,576-token context window and 65,536-token maximum output",
        "Google reports it ahead of Gemini 3.7 Flash on every published benchmark, including DeepSWE v1.1 73.7% vs 65.3% and Terminal-bench 2.1 89.4% vs 85.8%"
      ],
      "cautions": [
        "The listed price is promotional and ends 31 December 2026; on 1 January 2027 input and output both double, to $1.50 and $7.50, and cached input goes from $0.075 to $0.15",
        "Output pricing is billed inclusive of thinking tokens, so reasoning-heavy prompts cost more than the output length suggests",
        "Google’s published comparisons are against its own previous Flash release and were not independently reproduced here"
      ],
      "bestFor": [
        "High-volume agentic and coding workloads where per-token cost dominates the decision",
        "Multimodal pipelines that need audio, video and PDF input in one model"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.75,
        "outputPerMillionUsd": 3.75,
        "cachedInputPerMillionUsd": 0.075,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.75,
          "tier": "standard",
          "effectiveUntil": "2026-12-31",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 1.5,
          "tier": "standard",
          "effectiveFrom": "2027-01-01",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 3.75,
          "tier": "standard",
          "effectiveUntil": "2026-12-31",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 7.5,
          "tier": "standard",
          "effectiveFrom": "2027-01-01",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.075,
          "tier": "cached",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.375,
          "tier": "batch",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 1.875,
          "tier": "batch",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 1.35,
          "tier": "priority",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 6.75,
          "tier": "priority",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 95.4,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.8-flash_high"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 68.4,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.8-flash_high"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 69.7,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.8-flash_high"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1496,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 3.8 Flash (High)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1583,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 3.8 Flash (High)"
        }
      ],
      "officialSourceIds": [
        "google-gemini-38-flash-2026-09",
        "google-pricing"
      ]
    },
    {
      "id": "muse-spark-13",
      "providerId": "meta",
      "name": "Muse Spark 1.3",
      "family": "Muse Spark",
      "releaseDate": "2026-09-02",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1048576,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Meta tunes it for agentic workflows (multi-step tool use, browser control, and long-horizon tasks) with a 1,048,576-token context window",
        "Meta reports it finishing coding work with roughly 20% fewer tool calls and 25% fewer tokens than Muse Spark 1.2, which compounds with the low token price",
        "Behavioural changes aimed at real use: it asks clarifying questions on ambiguous prompts and confirms before consequential actions"
      ],
      "cautions": [
        "Meta’s own documentation states audio understanding is not fully supported and response quality may be degraded on requests containing audio",
        "The much cheaper contributor endpoint (about $0.10 / $0.20 per million) is priced that way in exchange for Meta training on your traffic. The standard tier is the rate recorded here.",
        "Meta’s scorecard has it trailing Claude Opus 5 on every broader agent-workflow benchmark it published, including GDPVal-AA v2, despite winning the coding rows"
      ],
      "bestFor": [
        "Tool-heavy agent loops where fewer calls per task matters as much as the per-token rate",
        "Coding agents at volume, where its coding scores and low price line up"
      ],
      "pricing": {
        "inputPerMillionUsd": 1.25,
        "outputPerMillionUsd": 4.25,
        "cachedInputPerMillionUsd": 0.15,
        "sourceId": "openrouter-models-api-2026-09"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 1.25,
          "tier": "standard",
          "sourceId": "openrouter-models-api-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 4.25,
          "tier": "standard",
          "sourceId": "openrouter-models-api-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.15,
          "tier": "cached",
          "sourceId": "openrouter-models-api-2026-09"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 74.4,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "muse-spark-1.3_xhigh"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1495,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Muse Spark 1.3 (Max)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1656,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Muse Spark 1.3 (Max)"
        }
      ],
      "officialSourceIds": [
        "meta-model-api-models-2026-09"
      ]
    },
    {
      "id": "qwen38-max",
      "providerId": "qwen",
      "name": "Qwen3.8-Max",
      "family": "Qwen3.8",
      "releaseDate": "2026-08-02",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": 131072,
      "modalities": [
        "text",
        "image",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "A 2.4-trillion-parameter mixture-of-experts model activating about 95 billion parameters per query, with a 1M-token context window and reasoning enabled by default",
        "Output is the cheapest of any frontier-tier text model here at $6 per million, an eighth of GPT-6 Astra and Claude Fable 5.1",
        "Accepts text, image and video input and returns text",
        "The 0902 post-training refresh (2 September 2026) targets coding and agent performance specifically"
      ],
      "cautions": [
        "Two releases share the name: the original 2 August 2026 model and the 0902 refresh, a separate id at the same price",
        "The recorded rates are the international (Singapore) price; the Global and Beijing listings charge $1.65 input and $4.951 output per 1M tokens",
        "Alibaba publishes no fixed cached-input rate for it"
      ],
      "bestFor": [
        "Output-heavy generation workloads where the $6 per million output rate is the deciding factor",
        "Agentic and coding evaluations that need a 1M-token window at mid-tier cost"
      ],
      "pricing": {
        "inputPerMillionUsd": 2,
        "outputPerMillionUsd": 6,
        "cachedInputPerMillionUsd": null,
        "sourceId": "alibaba-model-studio-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "standard",
          "sourceId": "alibaba-model-studio-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 6,
          "tier": "standard",
          "sourceId": "alibaba-model-studio-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "alibaba-model-studio-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 92.7,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "qwen3.8-max_xhigh"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 74.7,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "qwen3.8-max_xhigh"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 45.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "qwen3.8-max_xhigh"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1482,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Qwen3.8-Max"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1672,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Qwen3.8-Max"
        }
      ],
      "officialSourceIds": [
        "alibaba-model-studio-models-2026-09",
        "alibaba-model-studio-pricing",
        "alibaba-model-studio-new-models"
      ]
    },
    {
      "id": "kimi-k3",
      "providerId": "kimi",
      "name": "Kimi K3",
      "family": "Kimi K3",
      "releaseDate": "2026-07-16",
      "status": "active",
      "access": "both",
      "contextWindowTokens": 1048576,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "A 2.8-trillion-parameter mixture-of-experts model with about 104 billion active parameters, a 1M-token context window, and native vision",
        "Available both as a hosted API and as open weights: the checkpoint shipped 27 July 2026 under the Kimi K3 License as a native MXFP4 quantisation",
        "Introduces Kimi Delta Attention, a hybrid linear attention scheme, together with attention residuals"
      ],
      "cautions": [
        "Hosted rates match Claude Sonnet 5 exactly at $3 / $15 while offering a larger context window, so the comparison turns on capability rather than price",
        "Running the open weights shifts cost to infrastructure, which the token rates recorded here do not represent",
        "Moonshot’s self-reported results place it ahead of Claude Opus 4.8 and GPT-5.5 but behind Claude Fable 5 and GPT-5.6 Sol; none of this was reproduced independently here"
      ],
      "bestFor": [
        "Teams that want a frontier-scale model they can also self-host, with one architecture across both paths",
        "Long-context work at mid-tier hosted pricing"
      ],
      "pricing": {
        "inputPerMillionUsd": 3,
        "outputPerMillionUsd": 15,
        "cachedInputPerMillionUsd": 0.3,
        "sourceId": "openrouter-models-api-2026-09"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 3,
          "tier": "standard",
          "sourceId": "openrouter-models-api-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 15,
          "tier": "standard",
          "sourceId": "openrouter-models-api-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.3,
          "tier": "cached",
          "sourceId": "openrouter-models-api-2026-09"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 93.1,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "kimi-k3_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 72.2,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "kimi-k3_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 50.6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "kimi-k3_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1488,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Kimi K3 (Max)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1656,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Kimi K3 (Max)"
        }
      ],
      "officialSourceIds": [
        "kimi-platform-pricing-2026-09"
      ]
    },
    {
      "id": "kimi-k3-1-pro",
      "providerId": "kimi",
      "name": "Kimi K3.1 Pro",
      "family": "Kimi K3",
      "releaseDate": "2026-10-23",
      "status": "active",
      "access": "both",
      "apiModelId": "kimi-k3.1-pro",
      "contextWindowTokens": 1048576,
      "maxOutputTokens": 1048576,
      "modalities": [
        "text",
        "image",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "agentic-coding",
        "image-understanding",
        "video-understanding"
      ],
      "inputModalities": [
        "text",
        "image",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Moonshot's flagship for long-horizon coding and knowledge work: Kimi K3 post-trained with far more compute, a 2.8T-parameter mixture-of-experts (104B active) with a 1,048,576-token context and image and video input",
        "Cheaper than Kimi K3 at $2 input, $10 output and $0.20 cached input per 1M tokens, against K3's $3 / $15 / $0.30",
        "Moonshot reports 65.7% on Terminal-Bench 4.0, against 66.4% for Claude Opus 5.5 and 58.2% for GPT-6 Astra, and 94.2% on GPQA-Diamond",
        "Open weights on Hugging Face since November 2026, with serving recipes for vLLM, SGLang and TokenSpeed, and up to 1,048,576 output tokens through the API"
      ],
      "cautions": [
        "Thinking is always on: reasoning effort runs low, high or max, and max is the default",
        "The Highspeed endpoint, two to three times faster by Moonshot's account, costs twice as much at $4 input and $20 output per 1M tokens",
        "The weights carry Moonshot's own license: a model-as-a-service business above $20M in yearly revenue needs a separate agreement, and products above 100M monthly users or $20M monthly revenue must display the Kimi K3.1 Pro name",
        "Self-hosting takes a multi-node GPU cluster: the MXFP4 weights alone are about 1.56 TB",
        "Epoch AI and Arena Text have not scored it yet"
      ],
      "bestFor": [
        "Long-horizon coding and agent work at mid-tier prices",
        "Long documents, images and video in one 1M-token context"
      ],
      "pricing": {
        "inputPerMillionUsd": 2,
        "outputPerMillionUsd": 10,
        "cachedInputPerMillionUsd": 0.2,
        "sourceId": "kimi-platform-pricing-2026-09"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "standard",
          "sourceId": "kimi-platform-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 10,
          "tier": "standard",
          "sourceId": "kimi-platform-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.2,
          "tier": "cached",
          "sourceId": "kimi-platform-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "cache-write-5m",
          "sourceId": "kimi-platform-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 4,
          "tier": "cache-write-1h",
          "sourceId": "kimi-platform-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 4,
          "tier": "highspeed",
          "sourceId": "kimi-platform-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 20,
          "tier": "highspeed",
          "sourceId": "kimi-platform-pricing-2026-09"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1695,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Kimi K3.1 Pro (Max)"
        }
      ],
      "officialSourceIds": [
        "kimi-k3-1-pro-release",
        "kimi-k3-1-pro-quickstart",
        "kimi-platform-pricing-2026-09",
        "moonshot-kimi-k3-1-pro-weights"
      ]
    },
    {
      "id": "gemini-3-7-flash",
      "providerId": "google",
      "name": "Gemini 3.7 Flash",
      "family": "Gemini 3.7",
      "releaseDate": "2026-08-13",
      "status": "legacy",
      "deprecation": {
        "date": "2026-10-08",
        "replacementId": "gemini-38-flash",
        "sourceId": "google-gemini-changelog-2026-10-08"
      },
      "access": "api",
      "apiModelId": "gemini-3.7-flash",
      "contextWindowTokens": 1048576,
      "maxOutputTokens": 65536,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "coding",
        "agentic-coding",
        "instruction-following"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Google positions Gemini 3.7 Flash for higher-accuracy coding and agent workflows with multimodal inputs and a 1,048,576-token context window",
        "Google reports strong gains over 3.6 Flash in debugging and issue resolution, more diligent multi-step planning and tool calls, and closer instruction following",
        "Epoch AI measured 94.8% on GPQA Diamond and 71.6% on FrontierMath Tiers 1-3, ahead of 3.6 Flash on both"
      ],
      "cautions": [
        "Google deprecated it on 2026-10-08: Google AI Studio no longer offers it, and API requests to gemini-3.7-flash are routed to gemini-3.8-flash",
        "Google publishes the introductory API price through 2026-12-31; standard input and output prices increase on 2027-01-01.",
        "Not in the pinned LiveBench snapshot; its independent scores here come from Epoch AI's own GPQA Diamond, FrontierMath and SimpleQA Verified runs"
      ],
      "bestFor": [
        "Production coding and agent workflows that need multimodal context"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.75,
        "outputPerMillionUsd": 3.75,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-gemini-3-7-flash-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.75,
          "tier": "standard",
          "effectiveUntil": "2026-12-31",
          "sourceId": "google-gemini-3-7-flash-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": 3.75,
          "tier": "standard",
          "effectiveUntil": "2026-12-31",
          "sourceId": "google-gemini-3-7-flash-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-gemini-3-7-flash-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": 1.5,
          "tier": "standard",
          "effectiveFrom": "2027-01-01",
          "sourceId": "google-gemini-3-7-flash-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": 7.5,
          "tier": "standard",
          "effectiveFrom": "2027-01-01",
          "sourceId": "google-gemini-3-7-flash-release"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 94.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.7-flash_high"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 71.6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.7-flash_high"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 69.2,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.7-flash_high"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1490,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 3.7 Flash (High)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1590,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 3.7 Flash (High)"
        }
      ],
      "officialSourceIds": [
        "google-gemini-3-7-flash-model",
        "google-gemini-3-7-flash-release",
        "google-gemini-changelog-2026-10-08"
      ]
    },
    {
      "id": "gemini-3-1-pro-preview",
      "providerId": "google",
      "name": "Gemini 3.1 Pro Preview",
      "family": "Gemini 3.1",
      "releaseDate": "2026-02-19",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1048576,
      "maxOutputTokens": 65536,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "coding",
        "agentic-coding"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Google documents multimodal inputs, thinking, code execution, and function calling for Gemini 3.1 Pro Preview"
      ],
      "cautions": [
        "This preview endpoint may change before becoming stable; production use should account for its published prompt-length price tiers"
      ],
      "bestFor": [
        "Multimodal reasoning and agentic coding evaluations"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "standard-up-to-200k",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 12,
          "tier": "standard-up-to-200k",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.2,
          "tier": "cached-up-to-200k",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 4,
          "tier": "standard-over-200k",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 18,
          "tier": "standard-over-200k",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.4,
          "tier": "cached-over-200k",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 94.4,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.1-pro-preview_high"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 59.6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.1-pro-preview"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 73.5,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.1-pro-preview_high"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1487,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 3.1 Pro Preview"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1447,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 3.1 Pro Preview"
        }
      ],
      "officialSourceIds": [
        "google-gemini-3-1-pro-preview-model",
        "google-gemini-deprecations",
        "google-pricing"
      ]
    },
    {
      "id": "gemini-3-flash-preview",
      "providerId": "google",
      "name": "Gemini 3 Flash Preview",
      "family": "Gemini 3",
      "releaseDate": "2025-12-17",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1048576,
      "maxOutputTokens": 65536,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "coding",
        "agentic-coding"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Google documents multimodal inputs, thinking, computer use, code execution, and function calling"
      ],
      "cautions": [
        "This preview endpoint may change before becoming stable and the published price varies by input modality"
      ],
      "bestFor": [
        "Fast multimodal and agentic coding evaluations"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.5,
        "outputPerMillionUsd": 3,
        "cachedInputPerMillionUsd": 0.05,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.5,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 3,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.05,
          "tier": "cached",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 1,
          "tier": "standard-audio",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.1,
          "tier": "cached-audio",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 89.4,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3-flash-preview_high"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 51.2,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3-flash-preview"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 66.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3-flash-preview_high"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1473,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 3 Flash"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1440,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 3 Flash"
        }
      ],
      "officialSourceIds": [
        "google-gemini-3-flash-preview-model",
        "google-gemini-deprecations",
        "google-pricing"
      ]
    },
    {
      "id": "gemini-3-1-flash-lite",
      "providerId": "google",
      "name": "Gemini 3.1 Flash-Lite",
      "family": "Gemini 3.1",
      "releaseDate": "2026-05-07",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1048576,
      "maxOutputTokens": 65536,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "translation",
        "extraction"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Google documents low-latency multimodal processing for high-volume translation, extraction, and lightweight agentic tasks"
      ],
      "cautions": [
        "The published price for audio input differs from the listed text, image, and video input price",
        "Google's earliest published shutdown date is 2027-05-07; migrate to gemini-3.5-flash-lite."
      ],
      "bestFor": [
        "High-volume multimodal extraction and translation"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.25,
        "outputPerMillionUsd": 1.5,
        "cachedInputPerMillionUsd": 0.025,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.25,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 1.5,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.025,
          "tier": "cached",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.5,
          "tier": "standard-audio",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.05,
          "tier": "cached-audio",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 81.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.1-flash-lite_high"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 27.7,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-3.1-flash-lite_high"
        }
      ],
      "officialSourceIds": [
        "google-gemini-3-1-flash-lite-model",
        "google-gemini-deprecations",
        "google-pricing"
      ]
    },
    {
      "id": "gemini-3-5-live-translate",
      "providerId": "google",
      "name": "Gemini 3.5 Live Translate Preview",
      "family": "Gemini 3.5",
      "releaseDate": null,
      "status": "active",
      "access": "api",
      "contextWindowTokens": 131072,
      "maxOutputTokens": 65536,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "realtime-voice",
        "speech-translation"
      ],
      "inputModalities": [
        "audio"
      ],
      "outputModalities": [
        "text",
        "audio"
      ],
      "strengths": [
        "Google documents low-latency real-time speech-to-speech translation across 70+ languages"
      ],
      "cautions": [
        "This preview endpoint supports translation only, rather than tools or instructions, and may change before stable release"
      ],
      "bestFor": [
        "Realtime spoken-language translation"
      ],
      "pricing": {
        "inputPerMillionUsd": 3.5,
        "outputPerMillionUsd": 21,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 3.5,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 21,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing"
        },
        {
          "unit": "audio-minute",
          "usd": 0.0368,
          "tier": "effective-input-and-output",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "google-gemini-3-5-live-translate-model",
        "google-live-api-guide",
        "google-pricing"
      ]
    },
    {
      "id": "gemini-3-1-flash-live",
      "providerId": "google",
      "name": "Gemini 3.1 Flash Live Preview",
      "family": "Gemini 3.1",
      "releaseDate": "2026-03-26",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 131072,
      "maxOutputTokens": 65536,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "audio-understanding",
        "realtime-voice",
        "text-to-speech",
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "outputModalities": [
        "text",
        "audio"
      ],
      "strengths": [
        "Google documents a low-latency audio-to-audio model with multimodal awareness for realtime dialogue"
      ],
      "cautions": [
        "This preview endpoint may change before becoming stable and has input prices that vary by modality"
      ],
      "bestFor": [
        "Realtime voice applications with multimodal context"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing"
        },
        {
          "unit": "audio-minute",
          "usd": 0.005,
          "tier": "input",
          "sourceId": "google-pricing"
        },
        {
          "unit": "audio-minute",
          "usd": 0.018,
          "tier": "output",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.75,
          "tier": "standard-text",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 4.5,
          "tier": "standard-text",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 3,
          "tier": "standard-audio",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 12,
          "tier": "standard-audio",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 1,
          "tier": "standard-image-video",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "google-gemini-3-1-flash-live-model",
        "google-live-api-guide",
        "google-gemini-3-5-changelog",
        "google-pricing"
      ]
    },
    {
      "id": "gemini-3-1-flash-tts",
      "providerId": "google",
      "name": "Gemini 3.1 Flash TTS Preview",
      "family": "Gemini 3.1",
      "releaseDate": "2026-04-15",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 8192,
      "maxOutputTokens": 16384,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "text-to-speech"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "audio"
      ],
      "strengths": [
        "Google documents low-latency, controllable speech generation with expressive audio tags"
      ],
      "cautions": [
        "This preview endpoint may change before becoming stable and has more restrictive rate limits",
        "Google shuts this preview down on 17 November 2026 and names Gemini 3.8 Flash TTS or Gemini 3.8 Flash-Lite TTS as replacements"
      ],
      "bestFor": [
        "Controllable text-to-speech generation"
      ],
      "pricing": {
        "inputPerMillionUsd": 1,
        "outputPerMillionUsd": 20,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 1,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 20,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "google-gemini-3-1-flash-tts-model",
        "google-tts-guide",
        "google-gemini-3-5-changelog",
        "google-pricing",
        "google-gemini-3-preview-shutdowns"
      ]
    },
    {
      "id": "gemini-3-1-flash-lite-image",
      "providerId": "google",
      "name": "Gemini 3.1 Flash Lite Image",
      "family": "Gemini Image",
      "releaseDate": "2026-06-30",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 65536,
      "maxOutputTokens": 4096,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "image"
      ],
      "capabilities": [
        "image-generation",
        "image-editing",
        "image-to-image"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text",
        "image"
      ],
      "strengths": [
        "Google documents ultra-low-latency native image generation and editing optimized for 1K output"
      ],
      "cautions": [
        "Only 1K output is supported; Google documents 2K and 4K as unsupported"
      ],
      "bestFor": [
        "Low-latency 1K image generation and edits"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.25,
        "outputPerMillionUsd": 1.5,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.25,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 1.5,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing"
        },
        {
          "unit": "image",
          "usd": 0.0336,
          "tier": "standard-1k",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-image",
          "benchmarkSnapshotId": "arena-text-to-image-snapshot-2026-11-04",
          "score": 1249,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "Gemini 3.1 Flash-Lite Image (nano-banana-2-lite)"
        },
        {
          "benchmarkId": "arena-image-edit",
          "benchmarkSnapshotId": "arena-image-edit-snapshot-2026-11-04",
          "score": 1313,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "Gemini 3.1 Flash-Lite Image (nano-banana-2-lite)"
        }
      ],
      "officialSourceIds": [
        "google-gemini-3-1-flash-lite-image-model",
        "google-gemini-3-5-changelog",
        "google-pricing"
      ]
    },
    {
      "id": "gemini-3-pro-image",
      "providerId": "google",
      "name": "Gemini 3 Pro Image",
      "family": "Gemini Image",
      "releaseDate": "2026-05-28",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 65536,
      "maxOutputTokens": 32768,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "image"
      ],
      "capabilities": [
        "image-generation",
        "image-editing",
        "image-to-image"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text",
        "image"
      ],
      "strengths": [
        "Google documents native image generation and professional-grade image editing with image and text input"
      ],
      "cautions": [
        "Pricing differs by standard, batch, flex, and priority consumption"
      ],
      "bestFor": [
        "Professional image generation and editing"
      ],
      "pricing": {
        "inputPerMillionUsd": 2,
        "outputPerMillionUsd": 12,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 12,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing"
        },
        {
          "unit": "image",
          "usd": 0.134,
          "tier": "standard-1k-or-2k",
          "sourceId": "google-pricing"
        },
        {
          "unit": "image",
          "usd": 0.24,
          "tier": "standard-4k",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-image",
          "benchmarkSnapshotId": "arena-text-to-image-snapshot-2026-11-04",
          "score": 1246,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "Gemini 3 Pro Image (nano-banana-pro, 2K)"
        },
        {
          "benchmarkId": "arena-image-edit",
          "benchmarkSnapshotId": "arena-image-edit-snapshot-2026-11-04",
          "score": 1390,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "Gemini 3 Pro Image (nano-banana-pro, 2K)"
        }
      ],
      "officialSourceIds": [
        "google-gemini-3-pro-image-model",
        "google-gemini-deprecations",
        "google-gemini-3-5-changelog",
        "google-pricing"
      ]
    },
    {
      "id": "gemini-omni-flash",
      "providerId": "google",
      "name": "Gemini Omni Flash Preview",
      "family": "Gemini Omni",
      "releaseDate": "2026-06-30",
      "status": "legacy",
      "access": "api",
      "contextWindowTokens": 1048576,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "video"
      ],
      "capabilities": [
        "text-to-video",
        "image-to-video",
        "video-editing",
        "sound-generation"
      ],
      "inputModalities": [
        "text",
        "image",
        "video"
      ],
      "outputModalities": [
        "video",
        "audio"
      ],
      "strengths": [
        "Google documents conversational video generation and editing from text and images through the Interactions API"
      ],
      "cautions": [
        "Google deprecated this preview endpoint on 2026-09-30 and has since shut it down; Gemini Omni 1.2 Flash is the current generally available version",
        "This preview endpoint may change before becoming stable; it generates 3 to 10 second 720p video"
      ],
      "bestFor": [
        "Conversational video generation and editing"
      ],
      "pricing": {
        "inputPerMillionUsd": 1.5,
        "outputPerMillionUsd": 9,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 1.5,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 9,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing"
        },
        {
          "unit": "video-second",
          "usd": 0.1,
          "tier": "standard-720p",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-video",
          "benchmarkSnapshotId": "arena-text-to-video-snapshot-2026-11-05",
          "score": 1509,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-05",
          "sourceRow": "gemini-omni-flash"
        },
        {
          "benchmarkId": "arena-image-to-video",
          "benchmarkSnapshotId": "arena-image-to-video-snapshot-2026-11-05",
          "score": 1463,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-05",
          "sourceRow": "gemini-omni-flash"
        }
      ],
      "officialSourceIds": [
        "google-gemini-omni-flash-model",
        "google-omni-flash-guide",
        "google-gemini-3-5-changelog",
        "google-pricing"
      ]
    },
    {
      "id": "claude-opus-55",
      "providerId": "anthropic",
      "name": "Claude Opus 5.5",
      "family": "Claude Opus",
      "releaseDate": "2026-09-22",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Anthropic's recommended default for most workloads, taking that role from Claude Opus 5, at $4 per million input and $20 per million output",
        "1M-token context window with a 128,000-token maximum output, and adaptive thinking that is always on",
        "Cache reads are 0.05x base input ($0.20 per million) rather than the usual 0.1x, which matters most in agent loops that replay a large prefix"
      ],
      "cautions": [
        "The API effort default is medium here, not high as on Claude Fable 5.1 and Claude Sonnet 5.5, so a like-for-like comparison has to set effort explicitly",
        "The 4.7-generation tokenizer produces roughly 30% more tokens for the same text than Claude Sonnet 4.6 and earlier, so per-token rates do not compare directly across generations"
      ],
      "bestFor": [
        "Long-running agentic coding and knowledge work as the general-purpose default",
        "Cache-heavy agent loops where the 0.05x read multiplier compounds"
      ],
      "pricing": {
        "inputPerMillionUsd": 4,
        "outputPerMillionUsd": 20,
        "cachedInputPerMillionUsd": 0.2,
        "sourceId": "anthropic-pricing-2026-09"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 4,
          "tier": "standard",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 20,
          "tier": "standard",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.2,
          "tier": "cached",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 5,
          "tier": "cache-write-5m",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 8,
          "tier": "cache-write-1h",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "batch",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 10,
          "tier": "batch",
          "sourceId": "anthropic-pricing-2026-09"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 90.6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-opus-5-5_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 91.2,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-opus-5-5_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 72.2,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-opus-5-5_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1509,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude Opus 5.5 (High)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1810,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude Opus 5.5 (Max)"
        }
      ],
      "officialSourceIds": [
        "anthropic-models-overview-2026-09",
        "anthropic-pricing-2026-09"
      ]
    },
    {
      "id": "claude-sonnet-55",
      "providerId": "anthropic",
      "name": "Claude Sonnet 5.5",
      "family": "Claude Sonnet",
      "releaseDate": "2026-09-28",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Anthropic's speed-and-intelligence tier at $2 per million input and $10 per million output, with the same 1M-token window and 128,000-token output cap as Opus 5.5",
        "Cache reads at 0.05x base input ($0.10 per million), half the usual multiplier",
        "Keeps the high effort default, unlike Claude Opus 5.5"
      ],
      "cautions": [
        "Same list price as Claude Sonnet 5, so the gain is capability and cache economics rather than a cheaper rate card",
        "Anthropic positions Opus 5.5 rather than this as the default, so reach for Sonnet 5.5 when latency matters more than peak capability"
      ],
      "bestFor": [
        "Production workloads that need frontier-adjacent quality at a mid-tier rate",
        "Latency-sensitive agent steps inside a larger pipeline"
      ],
      "pricing": {
        "inputPerMillionUsd": 2,
        "outputPerMillionUsd": 10,
        "cachedInputPerMillionUsd": 0.1,
        "sourceId": "anthropic-pricing-2026-09"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "standard",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 10,
          "tier": "standard",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.1,
          "tier": "cached",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 2.5,
          "tier": "cache-write-5m",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 4,
          "tier": "cache-write-1h",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 1,
          "tier": "batch",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 5,
          "tier": "batch",
          "sourceId": "anthropic-pricing-2026-09"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 95.6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-sonnet-5-5_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 88.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-sonnet-5-5_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 46.5,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-sonnet-5-5_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1477,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude Sonnet 5.5 (xHigh)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1770,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude Sonnet 5.5 (xHigh)"
        }
      ],
      "officialSourceIds": [
        "anthropic-models-overview-2026-09",
        "anthropic-pricing-2026-09"
      ]
    },
    {
      "id": "claude-haiku-55",
      "providerId": "anthropic",
      "name": "Claude Haiku 5.5",
      "family": "Claude Haiku",
      "releaseDate": "2026-10-07",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "The cheapest model in this catalogue at $0.10 per million input and $0.50 per million output for prompts up to 100,000 tokens",
        "Carries the full 1M-token context window and a 128,000-token output cap, which earlier Haiku releases did not",
        "Cache reads at $0.01 per million make high-volume classification and routing nearly free on the input side"
      ],
      "cautions": [
        "Priced by prompt length, and it is a cliff rather than a taper: a request whose prompt passes 100,000 tokens is billed at $0.50 input and $2.50 output for the whole request, five times the short rate",
        "Cache reads and cache writes both count toward that 100,000-token measurement, so a large cached prefix can tip a request into the higher band on its own",
        "It is the only current Claude model whose 1M window is not priced uniformly; every other 4.6-and-later model bills a 900k request at the same rate as a 9k one"
      ],
      "bestFor": [
        "High-volume classification, extraction and routing under the 100,000-token line",
        "Cost-sensitive fan-out steps where a frontier model is overkill"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.1,
        "outputPerMillionUsd": 0.5,
        "cachedInputPerMillionUsd": 0.01,
        "sourceId": "anthropic-pricing-2026-09"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.1,
          "tier": "standard",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 0.5,
          "tier": "standard",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.01,
          "tier": "cached",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.125,
          "tier": "cache-write-5m",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.2,
          "tier": "cache-write-1h",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.5,
          "tier": "long-prompt-over-100k",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 2.5,
          "tier": "long-prompt-over-100k",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.05,
          "tier": "batch",
          "sourceId": "anthropic-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 0.25,
          "tier": "batch",
          "sourceId": "anthropic-pricing-2026-09"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 89.6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-haiku-5-5_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 75.1,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-haiku-5-5_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 23.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "claude-haiku-5-5_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1425,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude Haiku 5.5 (High)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1589,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Claude Haiku 5.5 (High)"
        }
      ],
      "officialSourceIds": [
        "anthropic-models-overview-2026-09",
        "anthropic-pricing-2026-09"
      ]
    },
    {
      "id": "gpt-6-sol",
      "providerId": "openai",
      "name": "GPT-6 Sol",
      "family": "GPT-6",
      "releaseDate": "2026-09-22",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1050000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "The mid tier of the GPT-6 family at $2 per million input and $10 per million output, a fifth of GPT-6 Astra on both",
        "Same 1,050,000-token context window and 128,000-token output cap as GPT-6 Astra"
      ],
      "cautions": [
        "Prompts over 272,000 input tokens are rebilled for the whole request at $4 input and $15 output, so crossing the line roughly doubles input cost",
        "GPT-6.1 Sol lists the same base rates with half the cached-input price, so there is little reason to pick this one for a cache-heavy workload"
      ],
      "bestFor": [
        "General production work that does not need Astra's capability or price",
        "Workloads kept under the 272,000-token threshold"
      ],
      "pricing": {
        "inputPerMillionUsd": 2,
        "outputPerMillionUsd": 10,
        "cachedInputPerMillionUsd": 0.2,
        "sourceId": "openai-api-pricing-2026-09"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "standard",
          "sourceId": "openai-api-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 10,
          "tier": "standard",
          "sourceId": "openai-api-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.2,
          "tier": "cached",
          "sourceId": "openai-api-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 4,
          "tier": "long-context-over-272k",
          "sourceId": "openai-api-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 15,
          "tier": "long-context-over-272k",
          "sourceId": "openai-api-pricing-2026-09"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 94.3,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-6-sol_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 89.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-6-sol_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 60.7,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-6-sol_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1458,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-6 Sol (Max)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1678,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-6 Sol (Max)"
        }
      ],
      "officialSourceIds": [
        "openai-api-pricing-2026-09",
        "openai-gpt-6-astra-model"
      ]
    },
    {
      "id": "gpt-61-sol",
      "providerId": "openai",
      "name": "GPT-6.1 Sol",
      "family": "GPT-6.1",
      "releaseDate": "2026-09-29",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1050000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Matches GPT-6 Sol's $2 / $10 base rates while halving cached input to $0.10 per million, the cheapest cache read of any OpenAI model here",
        "1,050,000-token context window with a 128,000-token maximum output"
      ],
      "cautions": [
        "The same 272,000-token long-context cliff applies, rebilling the full request at $4 input and $15 output",
        "Released a week after GPT-6 Sol at identical base pricing, so the two are easy to confuse on a rate card; the cached rate is the difference"
      ],
      "bestFor": [
        "Agent loops that replay a large cached prefix",
        "A drop-in alternative to GPT-6 Sol at the same base cost"
      ],
      "pricing": {
        "inputPerMillionUsd": 2,
        "outputPerMillionUsd": 10,
        "cachedInputPerMillionUsd": 0.1,
        "sourceId": "openai-api-pricing-2026-09"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "standard",
          "sourceId": "openai-api-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 10,
          "tier": "standard",
          "sourceId": "openai-api-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.1,
          "tier": "cached",
          "sourceId": "openai-api-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 4,
          "tier": "long-context-over-272k",
          "sourceId": "openai-api-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 15,
          "tier": "long-context-over-272k",
          "sourceId": "openai-api-pricing-2026-09"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 95.4,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-6.1-sol_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 93.7,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-6.1-sol_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 73.9,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-6.1-sol_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1481,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-6.1 Sol (Max)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1751,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-6.1 Sol (Max)"
        }
      ],
      "officialSourceIds": [
        "openai-api-pricing-2026-09",
        "openai-gpt-6-astra-model"
      ]
    },
    {
      "id": "gpt-61-terra",
      "providerId": "openai",
      "name": "GPT-6.1 Terra",
      "family": "GPT-6.1",
      "releaseDate": "2026-10-26",
      "status": "active",
      "access": "api",
      "apiModelId": "gpt-6.1-terra",
      "contextWindowTokens": 1050000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Half GPT-6.1 Sol's rates at $1 input and $5 output per 1M tokens, with cached input at $0.05; Batch and Flex halve them again",
        "The same 1.05M-token context and 128K maximum output as GPT-6.1 Sol, with image input and six reasoning effort levels from none to max",
        "Runs the Responses API tools, including computer use, web search, file search, code interpreter and MCP"
      ],
      "cautions": [
        "Prompts over 272,000 input tokens bill the whole request at $2 input and $7.50 output per 1M tokens",
        "Reads text and images only: audio and video input are not supported"
      ],
      "bestFor": [
        "Computer-use and tool-calling agents at half GPT-6.1 Sol's price",
        "High-volume reasoning work that stays under the 272K long-context threshold"
      ],
      "pricing": {
        "inputPerMillionUsd": 1,
        "outputPerMillionUsd": 5,
        "cachedInputPerMillionUsd": 0.05,
        "sourceId": "openai-gpt-6-1-terra-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 1,
          "tier": "standard",
          "sourceId": "openai-gpt-6-1-terra-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": 5,
          "tier": "standard",
          "sourceId": "openai-gpt-6-1-terra-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.05,
          "tier": "cached",
          "sourceId": "openai-gpt-6-1-terra-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "long-context-over-272k",
          "sourceId": "openai-gpt-6-1-terra-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": 7.5,
          "tier": "long-context-over-272k",
          "sourceId": "openai-gpt-6-1-terra-model"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 94.3,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-6.1-terra_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 90.9,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-6.1-terra_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 57.3,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-6.1-terra_max"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1673,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-6.1 Terra (Max)"
        }
      ],
      "officialSourceIds": [
        "openai-gpt-6-1-terra-model"
      ]
    },
    {
      "id": "gpt-6-luna",
      "providerId": "openai",
      "name": "GPT-6 Luna",
      "family": "GPT-6",
      "releaseDate": "2026-09-22",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 1050000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "OpenAI's budget tier at $0.10 per million input and $0.50 per million output, matching Claude Haiku 5.5 on both",
        "Keeps the full 1,050,000-token context window rather than cutting it down for the cheap tier",
        "The long-context surcharge starts at 272,000 tokens rather than Claude Haiku 5.5's 100,000, which is the more forgiving threshold of the two budget options"
      ],
      "cautions": [
        "Prompts over 272,000 tokens are rebilled at $0.20 input and $0.75 output for the whole request",
        "A twentieth of GPT-6 Sol's price reflects a capability gap, not a discount on the same model"
      ],
      "bestFor": [
        "High-volume summarisation, extraction and routing",
        "Long-context budget work, where the 272k threshold is more forgiving than Haiku 5.5's 100k"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.1,
        "outputPerMillionUsd": 0.5,
        "cachedInputPerMillionUsd": 0.01,
        "sourceId": "openai-api-pricing-2026-09"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.1,
          "tier": "standard",
          "sourceId": "openai-api-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 0.5,
          "tier": "standard",
          "sourceId": "openai-api-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.01,
          "tier": "cached",
          "sourceId": "openai-api-pricing-2026-09"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.2,
          "tier": "long-context-over-272k",
          "sourceId": "openai-api-pricing-2026-09"
        },
        {
          "unit": "million-output-tokens",
          "usd": 0.75,
          "tier": "long-context-over-272k",
          "sourceId": "openai-api-pricing-2026-09"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 90.5,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-6-luna_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 78.9,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-6-luna_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 41.4,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gpt-6-luna_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1441,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-6 Luna (Max)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1579,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GPT-6 Luna (Max)"
        }
      ],
      "officialSourceIds": [
        "openai-api-pricing-2026-09",
        "openai-gpt-6-astra-model"
      ]
    },
    {
      "id": "grok-47",
      "providerId": "xai",
      "name": "Grok 4.7",
      "family": "Grok 4",
      "releaseDate": "2026-09-21",
      "status": "active",
      "access": "api",
      "contextWindowTokens": 500000,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Holds Grok 4.6's $2 / $6 rates and 500,000-token context window",
        "Output at $6 per million undercuts every frontier-tier model here except the explicit budget tiers"
      ],
      "cautions": [
        "The long-prompt threshold is 200,000 tokens, the lowest of any model here, and reaching it doubles input and output for the entire request",
        "xAI's model page does not state a maximum output for Grok 4.7, so it is recorded as unknown rather than inferred",
        "That page does not explicitly list image input for this model either, so it is catalogued as text-only"
      ],
      "bestFor": [
        "Output-heavy generation under the 200,000-token prompt line",
        "A cheaper alternative to the frontier tiers where 500k of context is enough"
      ],
      "pricing": {
        "inputPerMillionUsd": 2,
        "outputPerMillionUsd": 6,
        "cachedInputPerMillionUsd": 0.5,
        "sourceId": "xai-models-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "standard",
          "sourceId": "xai-models-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 6,
          "tier": "standard",
          "sourceId": "xai-models-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.5,
          "tier": "cached",
          "sourceId": "xai-models-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 4,
          "tier": "long-prompt-over-200k",
          "sourceId": "xai-models-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 12,
          "tier": "long-prompt-over-200k",
          "sourceId": "xai-models-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 92.7,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "grok-4.7_xhigh"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 53,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "grok-4.7_xhigh"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 56,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "grok-4.7_xhigh"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1441,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Grok 4.7 (xHigh)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1637,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Grok 4.7 (xHigh)"
        }
      ],
      "officialSourceIds": [
        "xai-models-pricing"
      ]
    },
    {
      "id": "gemini-38-flash-tts",
      "providerId": "google",
      "name": "Gemini 3.8 Flash TTS",
      "family": "Gemini 3.8",
      "releaseDate": "2026-09-22",
      "status": "active",
      "access": "api",
      "apiModelId": "gemini-3.8-flash-tts",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "text-to-speech"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "audio"
      ],
      "strengths": [
        "Google's flagship expressive text-to-speech model, built for studio-grade voice fidelity, with custom voices and multi-speaker scene dialogue from text prompts",
        "Priced per token rather than per character: $0.50 per million text input tokens and $9.00 per million audio output tokens",
        "Shares the Gemini 3.8 Flash generation, so a text pipeline and its narration run on the same family"
      ],
      "cautions": [
        "Google doubles the rates on 2027-01-01, to $1.00 input and $18 output per 1M tokens (about $0.027 per minute)",
        "Audio output is billed in audio tokens, which do not convert one-for-one to text tokens, so a text-token budget does not predict this cost",
        "Google's pricing page lists no context window or output cap for this model, so both are recorded as unknown"
      ],
      "bestFor": [
        "Narration, audiobooks and product audio where voice quality matters more than cost",
        "Narration and voice output attached to an existing Gemini text pipeline",
        "Audio generation budgeted before the 2027 price step"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.5,
        "outputPerMillionUsd": 9,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.5,
          "tier": "standard",
          "effectiveUntil": "2026-12-31",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 9,
          "tier": "standard",
          "effectiveUntil": "2026-12-31",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 1,
          "tier": "standard",
          "effectiveFrom": "2027-01-01",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 18,
          "tier": "standard",
          "effectiveFrom": "2027-01-01",
          "sourceId": "google-pricing"
        },
        {
          "unit": "audio-minute",
          "usd": 0.0135,
          "tier": "standard",
          "effectiveUntil": "2026-12-31",
          "sourceId": "google-pricing"
        },
        {
          "unit": "audio-minute",
          "usd": 0.027,
          "tier": "standard",
          "effectiveFrom": "2027-01-01",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "google-gemini-3-8-tts-release",
        "google-gemini-3-5-changelog",
        "google-pricing",
        "google-gemini-38-flash-2026-09"
      ]
    },
    {
      "id": "deepseek-v41-flash",
      "providerId": "deepseek",
      "name": "DeepSeek-V4.1-Flash",
      "family": "DeepSeek-V4.1",
      "releaseDate": "2026-09-10",
      "status": "active",
      "access": "both",
      "apiModelId": "deepseek-flash",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": 384000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "A 552B-parameter mixture-of-experts model that activates 8B parameters for input and 16B for output, with native visual understanding and a 1M-token context window",
        "Released as MIT-licensed open weights on Hugging Face as well as on the DeepSeek API, so the same model can be hosted or self-run",
        "DeepSeek says it beats its own former flagship, DeepSeek-V4-Pro, and halves KV-cache memory needs; Epoch AI measured 89.8% on GPQA Diamond",
        "Updated on October 17, 2026 (build 1017, same API name and price): DeepSeek reports shorter outputs, faster task completion and gains on its own coding and agent tests"
      ],
      "cautions": [
        "The recorded API rates are the peak-hour prices; DeepSeek charges half outside 01:00-04:00 and 06:00-10:00 UTC on weekdays",
        "Running the open weights moves cost to your own infrastructure, which the token rates here do not represent",
        "Epoch AI and Arena WebDev scored build 1017, which the API serves; Arena Text has rated only the September build so far"
      ],
      "bestFor": [
        "High-volume agent and multimodal workloads where per-token cost dominates",
        "Teams that want a frontier-adjacent model they can also self-host under MIT terms"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.3,
        "outputPerMillionUsd": 1.2,
        "cachedInputPerMillionUsd": 0.006,
        "sourceId": "deepseek-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.3,
          "tier": "standard",
          "sourceId": "deepseek-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 1.2,
          "tier": "standard",
          "sourceId": "deepseek-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.006,
          "tier": "cached",
          "sourceId": "deepseek-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.15,
          "tier": "off-peak",
          "sourceId": "deepseek-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 0.6,
          "tier": "off-peak",
          "sourceId": "deepseek-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.003,
          "tier": "off-peak-cached",
          "sourceId": "deepseek-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 93.2,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "deepseek-v4.1-flash-1017_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 71.2,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "deepseek-v4.1-flash-1017_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 42.5,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "deepseek-v4.1-flash-1017_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1476,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "DeepSeek-V4.1-Flash-20260910 (Max)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1665,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "DeepSeek-V4.1-Flash-20261017 (Max)"
        }
      ],
      "officialSourceIds": [
        "deepseek-v4-1-flash-release",
        "deepseek-v4-1-flash-weights",
        "deepseek-pricing",
        "deepseek-v4-1-flash-1017-release",
        "deepseek-v4-1-flash-1017-weights"
      ]
    },
    {
      "id": "mistral-large-4",
      "providerId": "mistral",
      "name": "Mistral Large 4",
      "family": "Mistral Large 4",
      "releaseDate": "2026-10-06",
      "status": "active",
      "access": "both",
      "apiModelId": "mistral-large-4-0",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "A natively multimodal mixture-of-experts model with about 1 trillion parameters, 49 billion active per token, a 1M-token context window and text and image input",
        "Open weights since 29 October 2026 under the Mistral Ecosystem Open-Weight License, free to use, fine-tune and sell, with an NVFP4 checkpoint for Blackwell GPUs",
        "Cut to $0.70 input, $0.07 cached input and $2.10 output per 1M tokens with the open release, about half the preview list price",
        "Mistral reports 19.7% on Cognition FrontierCode 2.0 and 41% on Terminal-Bench 4.0, the best of the open-weight models it compared"
      ],
      "cautions": [
        "The API model is still labelled a public preview, and the open weights are a further-trained build of the October 6 preview",
        "Businesses that sell it as a hosted model service need a separate commercial licence from Mistral",
        "Running the open weights moves cost to your own GPUs, which the token rates here do not represent",
        "Epoch AI measured 22.4% on SimpleQA Verified for the open-weight build, low for its size, so check factual recall on your own questions"
      ],
      "bestFor": [
        "Self-hosted or European data-residency deployments that want a frontier-scale multimodal model",
        "Coding and security agents where open weights and a low API price both matter"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.7,
        "outputPerMillionUsd": 2.1,
        "cachedInputPerMillionUsd": 0.07,
        "sourceId": "mistral-large-4-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.7,
          "tier": "standard",
          "sourceId": "mistral-large-4-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": 2.1,
          "tier": "standard",
          "sourceId": "mistral-large-4-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.07,
          "tier": "cached",
          "sourceId": "mistral-large-4-model"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 90.2,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "mistral-large-4-0-oss_high"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 22.4,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "mistral-large-4-0-oss_high"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1432,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Mistral Large 4"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1544,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Mistral Large 4"
        }
      ],
      "officialSourceIds": [
        "mistral-large-4-release",
        "mistral-large-4-model",
        "mistral-large-4-open-weights",
        "mistral-large-4-weights"
      ]
    },
    {
      "id": "qwen38-omni-flash",
      "providerId": "qwen",
      "name": "Qwen3.8-Omni-Flash",
      "family": "Qwen3.8",
      "releaseDate": "2026-09-20",
      "status": "active",
      "access": "api",
      "apiModelId": "qwen3.8-omni-flash",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": 131072,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "multimodal",
        "audio"
      ],
      "capabilities": [
        "reasoning",
        "audio-understanding",
        "video-understanding",
        "image-understanding"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "An omnimodal model that reads text, images, audio and video in one 1M-token context and answers in text, with thinking, tool calling and web search",
        "Alibaba positions it for agents that plan, call tools and deliver finished work from audio-visual material, including up to an hour of combined audio and video",
        "Very low international API rates of $0.15 input and $0.47 output per 1M tokens"
      ],
      "cautions": [
        "Output is text only; for spoken replies Alibaba lists a separate realtime variant",
        "Alibaba's comparisons with Gemini 3.8 Flash are self-reported and have not been reproduced independently here",
        "The release date is the Alibaba Cloud announcement date, 2026-09-20"
      ],
      "bestFor": [
        "Long audio and video understanding at low cost",
        "Agent pipelines that must reason over meetings, calls or recorded media"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.15,
        "outputPerMillionUsd": 0.47,
        "cachedInputPerMillionUsd": 0.016,
        "sourceId": "alibaba-model-studio-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.15,
          "tier": "standard",
          "sourceId": "alibaba-model-studio-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 0.47,
          "tier": "standard",
          "sourceId": "alibaba-model-studio-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.016,
          "tier": "cached",
          "sourceId": "alibaba-model-studio-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "qwen-3-8-omni-flash-model",
        "qwen-3-8-omni-flash-blog",
        "alibaba-model-studio-pricing"
      ]
    },
    {
      "id": "gemini-nano-banana-2-1",
      "providerId": "google",
      "name": "Nano Banana 2.1",
      "family": "Nano Banana",
      "releaseDate": "2026-10-06",
      "status": "active",
      "access": "api",
      "apiModelId": "gemini-nano-banana-2.1",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "video"
      ],
      "taskFamilies": [
        "image",
        "multimodal"
      ],
      "capabilities": [
        "image-generation",
        "image-editing",
        "image-to-image"
      ],
      "inputModalities": [
        "text",
        "image",
        "video"
      ],
      "outputModalities": [
        "image",
        "text"
      ],
      "strengths": [
        "An update to Nano Banana 2 (Gemini 3.1 Flash Image) with better visual quality, prompt adherence, multi-turn character consistency and text rendering at Flash-tier speed",
        "Generates and edits at 1K, 2K and 4K, accepts text, image, video and PDF input, and fuses up to 14 reference images",
        "At $0.0336 per 1K image it costs about half of its predecessor's $0.067"
      ],
      "cautions": [
        "Google prices output by image tokens; the per-image figures are Google's own equivalents and batch is half price"
      ],
      "bestFor": [
        "High-volume product, marketing and social imagery where cost per image matters",
        "Iterative editing that must keep characters and objects consistent across turns"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 30,
          "tier": "standard-image-output",
          "sourceId": "google-pricing"
        },
        {
          "unit": "image",
          "usd": 0.0336,
          "tier": "standard-1k",
          "sourceId": "google-pricing"
        },
        {
          "unit": "image",
          "usd": 0.0504,
          "tier": "standard-2k",
          "sourceId": "google-pricing"
        },
        {
          "unit": "image",
          "usd": 0.113,
          "tier": "standard-4k",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-image",
          "benchmarkSnapshotId": "arena-text-to-image-snapshot-2026-11-04",
          "score": 1325,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "Gemini Nano Banana 2.1"
        },
        {
          "benchmarkId": "arena-image-edit",
          "benchmarkSnapshotId": "arena-image-edit-snapshot-2026-11-04",
          "score": 1427,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "Gemini Nano Banana 2.1"
        }
      ],
      "officialSourceIds": [
        "google-nano-banana-2-1-model",
        "google-gemini-3-5-changelog",
        "google-pricing"
      ]
    },
    {
      "id": "gemini-3-8-live",
      "providerId": "google",
      "name": "Gemini 3.8 Live",
      "family": "Gemini 3.8",
      "releaseDate": "2026-09-15",
      "status": "active",
      "access": "api",
      "apiModelId": "gemini-3.8-live",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "audio-understanding",
        "realtime-voice",
        "text-to-speech",
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "audio"
      ],
      "outputModalities": [
        "text",
        "audio"
      ],
      "strengths": [
        "Google's default speech-to-speech model for low-latency voice agents in the Gemini API and Google AI Studio",
        "A sibling, Gemini 3.8 Live Extended Thinking, trades latency for multi-step reasoning at the same published rates"
      ],
      "cautions": [
        "Audio is billed per token; the per-minute figures are Google's equivalents ($0.005 input, $0.018 output)",
        "No independent voice benchmark is recorded here"
      ],
      "bestFor": [
        "Customer-facing voice agents that need natural turn-taking"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing"
        },
        {
          "unit": "audio-minute",
          "usd": 0.005,
          "tier": "input",
          "sourceId": "google-pricing"
        },
        {
          "unit": "audio-minute",
          "usd": 0.018,
          "tier": "output",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.75,
          "tier": "standard-text",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 4.5,
          "tier": "standard-text",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 3,
          "tier": "standard-audio",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 12,
          "tier": "standard-audio",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "google-gemini-3-8-live-release",
        "google-gemini-3-5-changelog",
        "google-pricing"
      ]
    },
    {
      "id": "gemini-3-8-flash-lite-tts",
      "providerId": "google",
      "name": "Gemini 3.8 Flash-Lite TTS",
      "family": "Gemini 3.8",
      "releaseDate": "2026-09-22",
      "status": "active",
      "access": "api",
      "apiModelId": "gemini-3.8-flash-lite-tts",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "text-to-speech"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "audio"
      ],
      "strengths": [
        "A faster, cheaper text-to-speech model that replaces Gemini 3.1 Flash TTS Preview, with the same custom-voice and scene-dialogue controls",
        "Introductory output price of $6 per 1M audio tokens, about $0.009 per minute of speech, through 2026-12-31"
      ],
      "cautions": [
        "Google doubles the rates on 2027-01-01, to $1.00 input and $12 output per 1M tokens (about $0.018 per minute)"
      ],
      "bestFor": [
        "High-volume speech generation such as notifications, IVR prompts and accessibility audio"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.5,
        "outputPerMillionUsd": 6,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.5,
          "tier": "standard",
          "effectiveUntil": "2026-12-31",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 6,
          "tier": "standard",
          "effectiveUntil": "2026-12-31",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 1,
          "tier": "standard",
          "effectiveFrom": "2027-01-01",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 12,
          "tier": "standard",
          "effectiveFrom": "2027-01-01",
          "sourceId": "google-pricing"
        },
        {
          "unit": "audio-minute",
          "usd": 0.009,
          "tier": "standard",
          "effectiveUntil": "2026-12-31",
          "sourceId": "google-pricing"
        },
        {
          "unit": "audio-minute",
          "usd": 0.018,
          "tier": "standard",
          "effectiveFrom": "2027-01-01",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "google-gemini-3-8-tts-release",
        "google-gemini-3-5-changelog",
        "google-pricing"
      ]
    },
    {
      "id": "gpt-live-1",
      "providerId": "openai",
      "name": "GPT-Live 1",
      "family": "GPT-Live",
      "releaseDate": "2026-09-10",
      "status": "active",
      "access": "api",
      "apiModelId": "gpt-live-1",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "audio-understanding",
        "realtime-voice"
      ],
      "inputModalities": [
        "text",
        "audio"
      ],
      "outputModalities": [
        "text",
        "audio"
      ],
      "strengths": [
        "The full-duplex voice model from ChatGPT: it can listen and speak at the same time instead of strictly taking turns",
        "Hands reasoning and tool use to a backend model, so a voice front end can be paired with GPT-6-class reasoning",
        "Flat $0.05 per session minute, billed per second"
      ],
      "cautions": [
        "Backend reasoning and tool calls are billed separately at the configured model's rates, so real cost per minute is higher than the session rate",
        "OpenAI's task-success comparisons are self-reported and not reproduced independently here"
      ],
      "bestFor": [
        "Phone and in-app voice agents that need natural interruptions"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "openai-gpt-live-1-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "openai-gpt-live-1-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "openai-gpt-live-1-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "openai-gpt-live-1-model"
        },
        {
          "unit": "realtime-minute",
          "usd": 0.05,
          "tier": "standard",
          "sourceId": "openai-gpt-live-1-model"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "openai-gpt-live-1-model",
        "openai-api-changelog"
      ]
    },
    {
      "id": "gpt-image-2-5-sunburst",
      "providerId": "openai",
      "name": "GPT Image 2.5 Sunburst",
      "family": "GPT Image",
      "releaseDate": "2026-09-08",
      "status": "active",
      "access": "api",
      "apiModelId": "gpt-image-2.5-sunburst",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "image"
      ],
      "capabilities": [
        "image-generation",
        "image-editing",
        "image-to-image"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "image"
      ],
      "strengths": [
        "OpenAI's most capable image generation and editing model, with six quality settings from low to max",
        "Token rates match GPT Image 2: $5 per 1M text input tokens, $8 per 1M image input tokens and $30 per 1M image output tokens, half that in batch"
      ],
      "cautions": [
        "OpenAI publishes no per-image price, and its GPT Image 2 calculator does not estimate 2.5 token use, so cost per image has to be measured",
        "OpenAI followed it with GPT Image 2.6 Sunburst on November 4, 2026, which is faster at the same token rates, although Arena still rates 2.5 Sunburst higher"
      ],
      "bestFor": [
        "Hero imagery, detailed edits and text-heavy graphics where quality matters most"
      ],
      "pricing": {
        "inputPerMillionUsd": 5,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": 1.25,
        "sourceId": "openai-gpt-image-2-5-sunburst-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 5,
          "tier": "standard",
          "sourceId": "openai-gpt-image-2-5-sunburst-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "openai-gpt-image-2-5-sunburst-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 1.25,
          "tier": "cached",
          "sourceId": "openai-gpt-image-2-5-sunburst-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 8,
          "tier": "standard-image-input",
          "sourceId": "openai-gpt-image-2-5-sunburst-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": 30,
          "tier": "standard-image-output",
          "sourceId": "openai-gpt-image-2-5-sunburst-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": 15,
          "tier": "batch-image-output",
          "sourceId": "openai-gpt-image-2-5-sunburst-model"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-image",
          "benchmarkSnapshotId": "arena-text-to-image-snapshot-2026-11-04",
          "score": 1427,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "GPT-Image-2.5 Sunburst"
        },
        {
          "benchmarkId": "arena-image-edit",
          "benchmarkSnapshotId": "arena-image-edit-snapshot-2026-11-04",
          "score": 1517,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "GPT-Image-2.5 Sunburst"
        }
      ],
      "officialSourceIds": [
        "openai-gpt-image-2-5-sunburst-model",
        "openai-api-changelog"
      ]
    },
    {
      "id": "gpt-image-2-5-flare",
      "providerId": "openai",
      "name": "GPT Image 2.5 Flare",
      "family": "GPT Image",
      "releaseDate": "2026-09-08",
      "status": "active",
      "access": "api",
      "apiModelId": "gpt-image-2.5-flare",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "image"
      ],
      "capabilities": [
        "image-generation",
        "image-editing",
        "image-to-image"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "image"
      ],
      "strengths": [
        "OpenAI's fastest model for high-quality everyday image generation, with six quality settings from low to max",
        "Token rates match GPT Image 2: $5 per 1M text input tokens, $8 per 1M image input tokens and $30 per 1M image output tokens, half that in batch"
      ],
      "cautions": [
        "OpenAI publishes no per-image price, and its GPT Image 2 calculator does not estimate 2.5 token use, so cost per image has to be measured",
        "OpenAI followed it with GPT Image 2.6 Flare on November 4, 2026, which is faster at the same token rates, although Arena still rates 2.5 Flare higher"
      ],
      "bestFor": [
        "Fast, high-volume everyday image generation"
      ],
      "pricing": {
        "inputPerMillionUsd": 5,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": 1.25,
        "sourceId": "openai-gpt-image-2-5-flare-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 5,
          "tier": "standard",
          "sourceId": "openai-gpt-image-2-5-flare-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "openai-gpt-image-2-5-flare-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 1.25,
          "tier": "cached",
          "sourceId": "openai-gpt-image-2-5-flare-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 8,
          "tier": "standard-image-input",
          "sourceId": "openai-gpt-image-2-5-flare-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": 30,
          "tier": "standard-image-output",
          "sourceId": "openai-gpt-image-2-5-flare-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": 15,
          "tier": "batch-image-output",
          "sourceId": "openai-gpt-image-2-5-flare-model"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-image",
          "benchmarkSnapshotId": "arena-text-to-image-snapshot-2026-11-04",
          "score": 1399,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "GPT-Image-2.5 Flare"
        },
        {
          "benchmarkId": "arena-image-edit",
          "benchmarkSnapshotId": "arena-image-edit-snapshot-2026-11-04",
          "score": 1482,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "GPT-Image-2.5 Flare"
        }
      ],
      "officialSourceIds": [
        "openai-gpt-image-2-5-flare-model",
        "openai-api-changelog"
      ]
    },
    {
      "id": "grok-imagine-video-1-5-lite",
      "providerId": "xai",
      "name": "Grok Imagine Video 1.5 Lite",
      "family": "Grok Imagine",
      "releaseDate": null,
      "status": "active",
      "access": "api",
      "apiModelId": "grok-imagine-video-1.5-lite",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "video"
      ],
      "taskFamilies": [
        "video"
      ],
      "capabilities": [
        "text-to-video",
        "image-to-video"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "video"
      ],
      "strengths": [
        "xAI's lowest-cost video model, generating from text or an image at $0.020 per output second",
        "Supports the batch API"
      ],
      "cautions": [
        "xAI has not published a release date; trackers report early October 2026"
      ],
      "bestFor": [
        "Cheap drafts and simple clips before committing to a higher-end video model"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "xai-grok-imagine-video-1-5-lite-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "xai-grok-imagine-video-1-5-lite-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "xai-grok-imagine-video-1-5-lite-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "xai-grok-imagine-video-1-5-lite-model"
        },
        {
          "unit": "video-second",
          "usd": 0.02,
          "tier": "standard",
          "sourceId": "xai-grok-imagine-video-1-5-lite-model"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "xai-grok-imagine-video-1-5-lite-model"
      ]
    },
    {
      "id": "gemini-4-argon",
      "providerId": "google",
      "name": "Gemini 4 Argon",
      "family": "Gemini 4",
      "releaseDate": "2026-09-30",
      "status": "active",
      "access": "api",
      "apiModelId": "gemini-4-argon",
      "contextWindowTokens": 1048576,
      "maxOutputTokens": 1000000,
      "modalities": [
        "text",
        "image",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "coding",
        "agentic-coding",
        "image-understanding",
        "video-understanding",
        "document-understanding"
      ],
      "inputModalities": [
        "text",
        "image",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Google reports 77.9% on DeepSWE v1.1 for long-horizon software engineering and first place on AutomationBench at 51.3%",
        "Raises the output limit to 1M tokens, up from 64K on earlier Gemini models",
        "Ties for the top Arena Text rating in this catalogue at its announcement"
      ],
      "cautions": [
        "The $2 input and $10 output rates are introductory and end on November 30, 2026; from December 1 Google charges $4 and $20 per 1M tokens",
        "Not on the Gemini API free tier, and explicit context caching is not supported; implicit caching takes 95% off repeated input",
        "Answers longer than 64K tokens need background execution, and Google advises client timeouts of at least 15 minutes for hard prompts"
      ],
      "bestFor": [
        "Long-horizon coding and knowledge work once it reaches general API access"
      ],
      "pricing": {
        "inputPerMillionUsd": 2,
        "outputPerMillionUsd": 10,
        "cachedInputPerMillionUsd": 0.1,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "standard",
          "effectiveUntil": "2026-11-30",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 4,
          "tier": "standard",
          "effectiveFrom": "2026-12-01",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 10,
          "tier": "standard",
          "effectiveUntil": "2026-11-30",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 20,
          "tier": "standard",
          "effectiveFrom": "2026-12-01",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.1,
          "tier": "cached",
          "effectiveUntil": "2026-11-30",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.2,
          "tier": "cached",
          "effectiveFrom": "2026-12-01",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 96.4,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-4-argon_high"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 86,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-4-argon_high"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 76,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "gemini-4-argon_high"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1520,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 4 Argon (High)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1679,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 4 Argon (High)"
        }
      ],
      "officialSourceIds": [
        "google-gemini-4-argon-release",
        "google-gemini-4-argon-ga",
        "google-pricing"
      ]
    },
    {
      "id": "muse-wonder",
      "providerId": "meta",
      "name": "Muse Wonder",
      "family": "Muse Wonder",
      "releaseDate": "2026-10-12",
      "status": "active",
      "access": "api",
      "apiModelId": "muse-wonder-1.0",
      "contextWindowTokens": 1048576,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "coding",
        "agentic-coding",
        "instruction-following",
        "image-understanding",
        "video-understanding",
        "audio-understanding",
        "document-understanding"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Meta's most capable model, built for multi-file software changes and long-horizon agent runs with sub-agent delegation and parallel tool calls",
        "A 1,048,576-token context window with no long-context premium: a full window is billed at the same rate as a short prompt",
        "Uses the same endpoints as Muse Spark, so switching a workload is a model id change"
      ],
      "cautions": [
        "Meta lists it as a preview, and the Model API serves Standard mode only, with no discounted Contributor tier",
        "The higher-capability Muse Wonder Full is only available through Muse Code subscriptions, not per token"
      ],
      "bestFor": [
        "Multi-file coding changes and multi-hour agent runs",
        "Sub-agent orchestration where capability matters more than cost"
      ],
      "pricing": {
        "inputPerMillionUsd": 4,
        "outputPerMillionUsd": 20,
        "cachedInputPerMillionUsd": 0.2,
        "sourceId": "meta-model-api-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 4,
          "tier": "standard",
          "sourceId": "meta-model-api-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 20,
          "tier": "standard",
          "sourceId": "meta-model-api-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.2,
          "tier": "cached",
          "sourceId": "meta-model-api-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 82.1,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "muse-wonder-1.0_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1507,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Muse Wonder (Max)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1652,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Muse Wonder (Max)"
        }
      ],
      "officialSourceIds": [
        "meta-model-api-models-2026-09",
        "meta-muse-wonder-model-card",
        "meta-model-api-pricing"
      ]
    },
    {
      "id": "muse-glimmer",
      "providerId": "meta",
      "name": "Muse Glimmer",
      "family": "Muse Glimmer",
      "releaseDate": "2026-08-10",
      "status": "active",
      "access": "open-weights",
      "contextWindowTokens": 131072,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "coding",
        "agentic-coding",
        "image-understanding",
        "document-understanding"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "A 30-billion-parameter open-weight model distilled from Muse Spark, released under Apache 2.0 for local agent workflows",
        "Meta reports that 4-bit quantization brings the language model under 20 GB, enough for one consumer GPU or a Mac",
        "Meta reports 76.0 on SWE-Bench Verified and 75.5 on MCP Atlas at high reasoning"
      ],
      "cautions": [
        "Self-hosted only: Meta does not serve it over its Model API, so there is no first-party per-token price",
        "Audio is not supported and video is read as individual frames; the default context is 128K tokens"
      ],
      "bestFor": [
        "Local or on-device coding agents on a single GPU",
        "Fine-tuning an Apache 2.0 multimodal base"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "meta-muse-glimmer-docs"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "meta-muse-glimmer-docs"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "meta-muse-glimmer-docs"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "meta-muse-glimmer-docs"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1425,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Muse Glimmer"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1355,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Muse Glimmer"
        }
      ],
      "officialSourceIds": [
        "meta-muse-glimmer-docs",
        "meta-muse-glimmer-release"
      ]
    },
    {
      "id": "glm-5-3",
      "providerId": "zai",
      "name": "GLM-5.3",
      "family": "GLM-5",
      "releaseDate": "2026-08-18",
      "status": "active",
      "access": "both",
      "apiModelId": "glm-5.3",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning",
        "coding",
        "agentic-coding"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Zhipu reports a 50% gain over GLM-5.2 on its Z.ai Code Bench, from post-training on the same base model",
        "Open weights under the GLM-5.3 License, and an API that accepts OpenAI Chat Completions, OpenAI Responses and Anthropic Messages requests",
        "A 1M-token context window with up to 128K output tokens"
      ],
      "cautions": [
        "Text input only, and reasoning cannot be turned off: requests that disable thinking fail",
        "The licence requires a Z.ai security review before very large model-as-a-service operators use it commercially"
      ],
      "bestFor": [
        "Long-horizon agentic coding",
        "Self-hosted reasoning with permissive weights"
      ],
      "pricing": {
        "inputPerMillionUsd": 1.4,
        "outputPerMillionUsd": 4.4,
        "cachedInputPerMillionUsd": 0.26,
        "sourceId": "zai-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 1.4,
          "tier": "standard",
          "sourceId": "zai-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 4.4,
          "tier": "standard",
          "sourceId": "zai-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.26,
          "tier": "cached",
          "sourceId": "zai-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 90.9,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "glm-5.3_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 68.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "glm-5.3_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 41,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "glm-5.3_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1481,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GLM-5.3 (Max)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1623,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GLM-5.3 (Max)"
        }
      ],
      "officialSourceIds": [
        "zai-glm-5-3-docs",
        "zai-release-notes",
        "zai-pricing"
      ]
    },
    {
      "id": "glm-5-3-flash",
      "providerId": "zai",
      "name": "GLM-5.3-Flash",
      "family": "GLM-5",
      "releaseDate": "2026-08-26",
      "status": "active",
      "access": "both",
      "apiModelId": "glm-5.3-flash",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "coding",
        "agentic-coding",
        "image-understanding",
        "video-understanding",
        "document-understanding"
      ],
      "inputModalities": [
        "text",
        "image",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "The first natively multimodal GLM-5 model: it reads video, images, text and files and returns text",
        "320B total and 18B active parameters; Zhipu says hybrid sparse and linear attention cuts attention compute 3.01x and KV cache 4.44x against GLM-5.3",
        "MIT-licensed weights, and an API list price of $0.15 input and $0.50 output per 1M tokens"
      ],
      "cautions": [
        "Thinking cannot be disabled",
        "A faster GLM-5.3-FlashX is a separate model id at $0.37 input and $1.25 output per 1M tokens"
      ],
      "bestFor": [
        "Low-cost coding from screenshots or screen recordings",
        "Document and research agents on a budget"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.15,
        "outputPerMillionUsd": 0.5,
        "cachedInputPerMillionUsd": 0.03,
        "sourceId": "zai-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.15,
          "tier": "standard",
          "sourceId": "zai-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 0.5,
          "tier": "standard",
          "sourceId": "zai-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.03,
          "tier": "cached",
          "sourceId": "zai-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 90.2,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "glm-5.3-flash_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 55.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "glm-5.3-flash_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1477,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GLM-5.3-Flash"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1607,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "GLM-5.3-Flash"
        }
      ],
      "officialSourceIds": [
        "zai-glm-5-3-flash-docs",
        "zai-release-notes",
        "zai-pricing"
      ]
    },
    {
      "id": "deepseek-v4-pro",
      "providerId": "deepseek",
      "name": "DeepSeek-V4-Pro",
      "family": "DeepSeek-V4",
      "releaseDate": "2026-08-13",
      "status": "active",
      "access": "both",
      "apiModelId": "deepseek-v4-pro",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": 384000,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning",
        "coding",
        "agentic-coding"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "The 0813 GA release: DeepSeek reports 87.9 on Terminal Bench 2.1 and 62.7 on DeepSWE, up from 72.1 and 12.8 for the preview",
        "MIT-licensed weights, a 1M-token context and up to 384K output tokens",
        "Off-peak rates are half the listed price: $0.66 input and $1.98 output per 1M tokens"
      ],
      "cautions": [
        "The recorded rates are peak-hour prices (01:00 to 04:00 and 06:00 to 10:00 UTC on weekdays)",
        "No image input on the DeepSeek API, where the cheaper deepseek-flash has it",
        "DeepSeek had planned to retire it in September before extending service, so check its status before a long commitment"
      ],
      "bestFor": [
        "Agentic coding and terminal tasks",
        "Long-context jobs that need very long outputs"
      ],
      "pricing": {
        "inputPerMillionUsd": 1.32,
        "outputPerMillionUsd": 3.96,
        "cachedInputPerMillionUsd": 0.044,
        "sourceId": "deepseek-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 1.32,
          "tier": "standard",
          "sourceId": "deepseek-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 3.96,
          "tier": "standard",
          "sourceId": "deepseek-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.044,
          "tier": "cached",
          "sourceId": "deepseek-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.66,
          "tier": "off-peak",
          "sourceId": "deepseek-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 1.98,
          "tier": "off-peak",
          "sourceId": "deepseek-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.022,
          "tier": "off-peak-cached",
          "sourceId": "deepseek-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 91.7,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "deepseek-v4-pro-0813_max"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 64.6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "deepseek-v4-pro-0813_max"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 52.9,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "deepseek-v4-pro-0813_max"
        },
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1465,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "DeepSeek-V4-Pro (High, 20260813)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1582,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "DeepSeek-V4-Pro (High, 20260813)"
        }
      ],
      "officialSourceIds": [
        "deepseek-v4-pro-0813-release",
        "deepseek-v4-pro-0813-weights",
        "deepseek-pricing"
      ]
    },
    {
      "id": "mimo-v2-6-pro",
      "providerId": "xiaomi",
      "name": "MiMo-V2.6-Pro",
      "family": "MiMo-V2.6",
      "releaseDate": "2026-09-22",
      "status": "active",
      "access": "both",
      "apiModelId": "mimo-v2.6-pro",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "agentic-coding",
        "image-understanding",
        "video-understanding",
        "audio-understanding"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Xiaomi reports 89.9 on Terminal Bench 2.1 and 71.9 on DeepSWE v1.1",
        "A 1.02T-parameter mixture-of-experts model with 42B active that takes text, image, video and audio in one model with a 1M-token context",
        "MIT-licensed weights, and $0.435 input and $0.87 output per 1M tokens on the overseas API, halved on the Batch API"
      ],
      "cautions": [
        "The weights ship as two checkpoints (RL and MOPD), and Xiaomi does not say which one the API serves",
        "The mimo-v2.6-pro-ultraspeed id costs ten times as much and has no Batch API"
      ],
      "bestFor": [
        "Long-horizon agent and coding work across modalities",
        "High-volume batch workloads"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.435,
        "outputPerMillionUsd": 0.87,
        "cachedInputPerMillionUsd": 0.0036,
        "sourceId": "xiaomi-mimo-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.435,
          "tier": "standard",
          "sourceId": "xiaomi-mimo-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 0.87,
          "tier": "standard",
          "sourceId": "xiaomi-mimo-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.0036,
          "tier": "cached",
          "sourceId": "xiaomi-mimo-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.2175,
          "tier": "batch",
          "sourceId": "xiaomi-mimo-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 0.435,
          "tier": "batch",
          "sourceId": "xiaomi-mimo-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1480,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "MiMo-V2.6-Pro"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1625,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "MiMo-V2.6-Pro"
        }
      ],
      "officialSourceIds": [
        "xiaomi-mimo-model-updates",
        "mimo-v2-6-pro-weights",
        "xiaomi-mimo-pricing"
      ]
    },
    {
      "id": "mimo-v2-6-flash",
      "providerId": "xiaomi",
      "name": "MiMo-V2.6-Flash",
      "family": "MiMo-V2.6",
      "releaseDate": "2026-09-22",
      "status": "active",
      "access": "both",
      "apiModelId": "mimo-v2.6-flash",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "agentic-coding",
        "image-understanding",
        "video-understanding",
        "audio-understanding"
      ],
      "inputModalities": [
        "text",
        "image",
        "audio",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Xiaomi reports 87.6 on Terminal Bench 2.1, close to the Pro model at a third of the price",
        "A 309B-parameter mixture-of-experts model with 15B active, taking text, image, video and audio with a 1M-token context",
        "$0.14 input and $0.28 output per 1M tokens on the overseas API, halved on the Batch API"
      ],
      "cautions": [
        "Xiaomi's own table puts it well behind the Pro model on exploitation tasks (ExploitBench 25.3 against 47.9)",
        "The API checkpoint (RL or MOPD weights) is not identified"
      ],
      "bestFor": [
        "High-frequency multimodal calls",
        "Low-cost agent workloads"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.14,
        "outputPerMillionUsd": 0.28,
        "cachedInputPerMillionUsd": 0.0028,
        "sourceId": "xiaomi-mimo-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.14,
          "tier": "standard",
          "sourceId": "xiaomi-mimo-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 0.28,
          "tier": "standard",
          "sourceId": "xiaomi-mimo-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.0028,
          "tier": "cached",
          "sourceId": "xiaomi-mimo-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.07,
          "tier": "batch",
          "sourceId": "xiaomi-mimo-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 0.14,
          "tier": "batch",
          "sourceId": "xiaomi-mimo-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1448,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "MiMo-V2.6-Flash"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1637,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "MiMo-V2.6-Flash"
        }
      ],
      "officialSourceIds": [
        "xiaomi-mimo-model-updates",
        "mimo-v2-6-flash-weights",
        "xiaomi-mimo-pricing"
      ]
    },
    {
      "id": "qwen3-8-flash-next",
      "providerId": "qwen",
      "name": "Qwen3.8-Flash-Next",
      "family": "Qwen3.8",
      "releaseDate": "2026-08-26",
      "status": "active",
      "access": "open-weights",
      "contextWindowTokens": 262144,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "coding",
        "agentic-coding",
        "instruction-following",
        "image-understanding",
        "video-understanding"
      ],
      "inputModalities": [
        "text",
        "image",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "An experimental preview of the architecture Alibaba plans for Qwen4: 125B parameters with 6B active, plus a 51B n-gram embedding table",
        "Alibaba reports 62.5 on SWE-bench Pro and 91.7 on GPQA Diamond, ahead of the larger Qwen3.7-Plus in its own table",
        "Alibaba says training cost about one ninth of Qwen3.7-Plus"
      ],
      "cautions": [
        "Alibaba calls it experimental; the hosted qwen3.8-flash API is a different, production model",
        "The Qwen Community License requires a separate licence before a model-as-a-service business uses it",
        "Native context is 262,144 tokens; 1M needs YaRN scaling"
      ],
      "bestFor": [
        "Self-hosted agentic coding on limited accelerator memory",
        "Research on the Qwen4 architecture"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "qwen-3-8-flash-next-weights"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "qwen-3-8-flash-next-weights"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "qwen-3-8-flash-next-weights"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "qwen-3-8-flash-next-weights"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1628,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Qwen3.8-Flash-Next"
        }
      ],
      "officialSourceIds": [
        "qwen-3-8-flash-next-weights"
      ]
    },
    {
      "id": "nemotron-3-5-lightning",
      "providerId": "nvidia",
      "name": "Nemotron 3.5 Lightning",
      "family": "Nemotron 3.5",
      "releaseDate": "2026-08-11",
      "status": "active",
      "access": "open-weights",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning",
        "agentic-coding"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "A 30B mixture-of-experts model with 3B active parameters; NVIDIA reports up to 4x the output speed of similar-sized models",
        "Weights, training data and recipes are released under the OpenMDW-1.1 licence, in BF16 and NVFP4",
        "Context of up to 1M tokens, or 256K on a single H100"
      ],
      "cautions": [
        "On NVIDIA's own table it trails Qwen 3.6 35B A3B on SWE-bench Verified (51.56 against 70.12)",
        "NVIDIA sells no per-token API for it; hosted prices come from third parties"
      ],
      "bestFor": [
        "Fast, low-cost agent subtasks such as tool calls",
        "Self-hosted agents on one GPU"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "nvidia-nemotron-3-5-lightning-weights"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "nvidia-nemotron-3-5-lightning-weights"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "nvidia-nemotron-3-5-lightning-weights"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "nvidia-nemotron-3-5-lightning-weights"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1353,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "NVIDIA-Nemotron-3.5-Lightning-30B-A3B (NVFP4)"
        }
      ],
      "officialSourceIds": [
        "nvidia-nemotron-3-5-lightning-weights",
        "nvidia-nemotron-3-5-lightning-release"
      ]
    },
    {
      "id": "granite-4-2-30b",
      "providerId": "ibm",
      "name": "Granite 4.2 30B",
      "family": "Granite 4.2",
      "releaseDate": "2026-08-25",
      "status": "active",
      "access": "open-weights",
      "contextWindowTokens": 128000,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning",
        "coding"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "A 30B dense reasoning model under Apache 2.0; IBM reports 57 on SWE-Bench Verified and 89.17 on AIME25",
        "Thinking, non-thinking and low-effort modes, with reasoning-augmented tool calling",
        "A native 128K context, extendable to 512K"
      ],
      "cautions": [
        "IBM says it is fine-tuned mainly on English instruction data, so other languages may be weaker",
        "IBM publishes no per-token API price for Granite 4.2"
      ],
      "bestFor": [
        "Self-hosted enterprise agents that need a permissive licence"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "ibm-granite-4-2-30b-weights"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "ibm-granite-4-2-30b-weights"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "ibm-granite-4-2-30b-weights"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "ibm-granite-4-2-30b-weights"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1340,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Granite 4.2 30B"
        }
      ],
      "officialSourceIds": [
        "ibm-granite-4-2-30b-weights",
        "ibm-granite-4-2-docs"
      ]
    },
    {
      "id": "mercury-2-5",
      "providerId": "inception",
      "name": "Mercury 2.5",
      "family": "Mercury",
      "releaseDate": "2026-09-08",
      "status": "active",
      "access": "api",
      "apiModelId": "mercury-2.5",
      "contextWindowTokens": 260000,
      "maxOutputTokens": 65536,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "A diffusion language model that refines many tokens in parallel; Inception reports 1,107 tokens per second on widely available NVIDIA GPUs",
        "Tunable reasoning, parallel tool calls and schema-aligned JSON",
        "A 260K-token context with up to 65,536 output tokens"
      ],
      "cautions": [
        "Inception publishes no benchmark scores for it, and no independent score is recorded here",
        "The recorded rates are list prices; an 80% launch discount was running with no published end date"
      ],
      "bestFor": [
        "Latency-critical agent loops and voice agents",
        "High-throughput, low-cost text generation"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.2,
        "outputPerMillionUsd": 0.75,
        "cachedInputPerMillionUsd": 0.02,
        "sourceId": "inception-models"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.2,
          "tier": "standard",
          "sourceId": "inception-models"
        },
        {
          "unit": "million-output-tokens",
          "usd": 0.75,
          "tier": "standard",
          "sourceId": "inception-models"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.02,
          "tier": "cached",
          "sourceId": "inception-models"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "inception-models",
        "inception-mercury-2-5-release"
      ]
    },
    {
      "id": "solar-pro-4",
      "providerId": "upstage",
      "name": "Solar Pro 4",
      "family": "Solar Pro",
      "releaseDate": "2026-08-06",
      "status": "active",
      "access": "api",
      "apiModelId": "solar-pro4-260806",
      "contextWindowTokens": 512000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning",
        "coding",
        "agentic-coding",
        "document-understanding"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Upstage's agentic flagship for tool calling, terminal tasks and long-document reasoning, with a 512K context and up to 128K output tokens",
        "Upstage reports 70.6 on SWE-Bench Verified with OpenHands in its own evaluation",
        "Supports English, Korean and Japanese"
      ],
      "cautions": [
        "A closed model: Upstage does not disclose its parameter count",
        "Several of its published benchmark figures are in-house evaluations"
      ],
      "bestFor": [
        "Korean and Japanese enterprise workloads",
        "Multi-step agent work with tool calls"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.3,
        "outputPerMillionUsd": 1.2,
        "cachedInputPerMillionUsd": 0.06,
        "sourceId": "upstage-solar-pro-4-docs"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.3,
          "tier": "standard",
          "sourceId": "upstage-solar-pro-4-docs"
        },
        {
          "unit": "million-output-tokens",
          "usd": 1.2,
          "tier": "standard",
          "sourceId": "upstage-solar-pro-4-docs"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.06,
          "tier": "cached",
          "sourceId": "upstage-solar-pro-4-docs"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1388,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Solar Pro 4"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1370,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Solar Pro 4"
        }
      ],
      "officialSourceIds": [
        "upstage-solar-pro-4-docs"
      ]
    },
    {
      "id": "step-5-preview",
      "providerId": "stepfun",
      "name": "Step 5 Preview",
      "family": "Step 5",
      "releaseDate": null,
      "status": "active",
      "access": "api",
      "apiModelId": "step-5-preview",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": 64000,
      "modalities": [
        "text",
        "image",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "coding",
        "agentic-coding",
        "image-understanding",
        "document-understanding"
      ],
      "inputModalities": [
        "text",
        "image",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "StepFun's flagship for agentic work, with particular strength claimed in software engineering and finance",
        "Text, image and video input within a 1M-token context, with low, medium and high reasoning effort",
        "Tool calling, JSON Schema output and prompt caching"
      ],
      "cautions": [
        "A preview: StepFun publishes no release date, parameter count or benchmark scores for it",
        "Output tokens are billed including the reasoning trace, and output is capped at 64K tokens"
      ],
      "bestFor": [
        "Long-context agentic coding and finance work"
      ],
      "pricing": {
        "inputPerMillionUsd": 1,
        "outputPerMillionUsd": 2.7,
        "cachedInputPerMillionUsd": 0.1,
        "sourceId": "stepfun-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 1,
          "tier": "standard",
          "sourceId": "stepfun-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 2.7,
          "tier": "standard",
          "sourceId": "stepfun-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.1,
          "tier": "cached",
          "sourceId": "stepfun-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1448,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Step 5 Preview"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1571,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Step 5 Preview (High)"
        }
      ],
      "officialSourceIds": [
        "stepfun-step-5-preview-docs",
        "stepfun-pricing"
      ]
    },
    {
      "id": "gpt-realtime-2",
      "providerId": "openai",
      "name": "GPT-Realtime-2",
      "family": "GPT Realtime",
      "releaseDate": "2026-05-07",
      "status": "active",
      "access": "api",
      "apiModelId": "gpt-realtime-2",
      "contextWindowTokens": 128000,
      "maxOutputTokens": 32000,
      "modalities": [
        "text",
        "audio",
        "image"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "realtime-voice",
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "audio",
        "image"
      ],
      "outputModalities": [
        "text",
        "audio"
      ],
      "strengths": [
        "OpenAI's most capable realtime voice model, with configurable reasoning for speech-to-speech agents",
        "Text, audio and image input in one Realtime session, with function calling and prompt caching"
      ],
      "cautions": [
        "Available only on the Realtime endpoint, not Chat Completions, Responses or Batch",
        "Audio costs $32 input and $64 output per 1M tokens, eight and four times its text rates",
        "Higher reasoning effort adds latency and output tokens"
      ],
      "bestFor": [
        "Speech-to-speech voice agents with tool calls"
      ],
      "pricing": {
        "inputPerMillionUsd": 4,
        "outputPerMillionUsd": 24,
        "cachedInputPerMillionUsd": 0.4,
        "sourceId": "openai-gpt-realtime-2-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 4,
          "tier": "standard",
          "sourceId": "openai-gpt-realtime-2-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": 24,
          "tier": "standard",
          "sourceId": "openai-gpt-realtime-2-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.4,
          "tier": "cached",
          "sourceId": "openai-gpt-realtime-2-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 32,
          "tier": "audio-input",
          "sourceId": "openai-gpt-realtime-2-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": 64,
          "tier": "audio-output",
          "sourceId": "openai-gpt-realtime-2-model"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "openai-gpt-realtime-2-model",
        "openai-api-changelog"
      ]
    },
    {
      "id": "gpt-transcribe",
      "providerId": "openai",
      "name": "GPT Transcribe",
      "family": "GPT Transcribe",
      "releaseDate": "2026-07-28",
      "status": "active",
      "access": "api",
      "apiModelId": "gpt-transcribe",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "audio",
        "text"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "speech-to-text"
      ],
      "inputModalities": [
        "audio",
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "OpenAI's high-accuracy speech-to-text model for files and committed Realtime turns, priced at $0.0045 per minute",
        "Accepts free-form context, keyword hints and several expected languages for domain terms and code-switching",
        "The model OpenAI points users to as gpt-4o-transcribe and whisper-1 are retired"
      ],
      "cautions": [
        "Live microphone audio goes to a separate model, gpt-live-transcribe, at $0.017 per minute",
        "OpenAI's model page does not mention speaker diarization"
      ],
      "bestFor": [
        "Transcribing recorded audio files"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "openai-gpt-transcribe-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "openai-gpt-transcribe-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "openai-gpt-transcribe-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "openai-gpt-transcribe-model"
        },
        {
          "unit": "audio-minute",
          "usd": 0.0045,
          "tier": "standard",
          "sourceId": "openai-gpt-transcribe-model"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "openai-gpt-transcribe-model",
        "openai-api-changelog"
      ]
    },
    {
      "id": "eleven-v4",
      "providerId": "elevenlabs",
      "name": "Eleven v4",
      "family": "Eleven TTS",
      "releaseDate": "2026-09-28",
      "status": "active",
      "access": "api",
      "apiModelId": "eleven_v4",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "text-to-speech"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "audio"
      ],
      "strengths": [
        "ElevenLabs reports more than 90 languages and natural multi-speaker dialogue",
        "Up to 10,000 characters per request, which ElevenLabs puts at about 10 minutes of audio",
        "Same $0.08 per 1,000 characters as Eleven v3, which it replaces"
      ],
      "cautions": [
        "ElevenLabs' blind-test preference claims are its own; the independent rating is recorded below",
        "For low latency ElevenLabs points to Eleven v4 Turbo"
      ],
      "bestFor": [
        "Expressive narration and dialogue",
        "Multilingual voice content"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "elevenlabs-api-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "elevenlabs-api-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "elevenlabs-api-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "elevenlabs-api-pricing"
        },
        {
          "unit": "character",
          "usd": 0.00008,
          "tier": "standard",
          "sourceId": "elevenlabs-api-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "specialist-elevenlabs-models",
        "elevenlabs-eleven-v4-release",
        "elevenlabs-api-pricing"
      ]
    },
    {
      "id": "eleven-v4-turbo",
      "providerId": "elevenlabs",
      "name": "Eleven v4 Turbo",
      "family": "Eleven TTS",
      "releaseDate": "2026-09-28",
      "status": "active",
      "access": "api",
      "apiModelId": "eleven_v4_turbo",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "text-to-speech"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "audio"
      ],
      "strengths": [
        "ElevenLabs reports a median inference latency of about 100 ms and time to first speech of about 150 ms",
        "More than 90 languages and audio tags for fine-grained delivery control",
        "Half the per-character price of Eleven v4"
      ],
      "cautions": [
        "ElevenLabs' latency figures exclude application and network latency"
      ],
      "bestFor": [
        "Realtime voice agents",
        "Low-latency streaming speech"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "elevenlabs-api-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "elevenlabs-api-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "elevenlabs-api-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "elevenlabs-api-pricing"
        },
        {
          "unit": "character",
          "usd": 0.00004,
          "tier": "standard",
          "sourceId": "elevenlabs-api-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "specialist-elevenlabs-models",
        "elevenlabs-eleven-v4-release",
        "elevenlabs-api-pricing"
      ]
    },
    {
      "id": "grok-voice-transcribe-2",
      "providerId": "xai",
      "name": "Grok Voice Transcribe 2.0",
      "family": "Grok Voice",
      "releaseDate": "2026-09-18",
      "status": "active",
      "access": "api",
      "apiModelId": "grok-voice-transcribe-2.0",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "audio",
        "text"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "speech-to-text",
        "diarization"
      ],
      "inputModalities": [
        "audio"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "xAI reports it is twice as accurate as Grok Voice Transcribe 1.0 at the same price",
        "38+ languages with diarization, word timestamps, multichannel audio and key-term biasing",
        "Batch over REST at $0.10 per hour of audio, or streaming over WebSocket at $0.20 per hour"
      ],
      "cautions": [
        "xAI's per-model page leaves pricing blank; the rates come from its models table and launch post"
      ],
      "bestFor": [
        "Low-cost batch transcription",
        "Streaming transcription for voice apps"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "xai-developer-models"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "xai-developer-models"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "xai-developer-models"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "xai-developer-models"
        },
        {
          "unit": "audio-minute",
          "usd": 0.001667,
          "tier": "standard",
          "sourceId": "xai-developer-models"
        },
        {
          "unit": "audio-minute",
          "usd": 0.003333,
          "tier": "streaming",
          "sourceId": "xai-developer-models"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "xai-grok-voice-transcribe-2-model",
        "xai-grok-voice-transcribe-2-release",
        "xai-developer-models"
      ]
    },
    {
      "id": "mai-transcribe-2",
      "providerId": "microsoft",
      "name": "MAI-Transcribe-2",
      "family": "MAI-Transcribe",
      "releaseDate": "2026-09-03",
      "status": "active",
      "access": "api",
      "apiModelId": "MAI-Transcribe-2",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "audio",
        "text"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "speech-to-text",
        "diarization"
      ],
      "inputModalities": [
        "audio"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Microsoft reports first place on FLEURS across 60 languages with a 5.2% average word error rate",
        "Diarization, word timestamps, keyword biasing, language identification and code-switching",
        "$0.10 per hour of audio at launch"
      ],
      "cautions": [
        "Azure Speech lists it as a public preview with no SLA, not recommended for production",
        "The $0.10 per hour price is a limited-time offer until 2026-12-31, and the later price is not published"
      ],
      "bestFor": [
        "Low-cost multilingual batch transcription",
        "Meeting and call transcripts with speaker labels"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "microsoft-mai-transcribe-2-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "microsoft-mai-transcribe-2-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "microsoft-mai-transcribe-2-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "microsoft-mai-transcribe-2-release"
        },
        {
          "unit": "audio-minute",
          "usd": 0.001667,
          "tier": "standard",
          "effectiveUntil": "2026-12-31",
          "sourceId": "microsoft-mai-transcribe-2-release"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "microsoft-mai-transcribe-docs",
        "microsoft-mai-transcribe-2-release"
      ]
    },
    {
      "id": "sonic-3-6",
      "providerId": "cartesia",
      "name": "Sonic 3.6",
      "family": "Sonic",
      "releaseDate": "2026-08-27",
      "status": "active",
      "access": "api",
      "apiModelId": "sonic-3.6",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "text-to-speech"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "audio"
      ],
      "strengths": [
        "Cartesia reports 44 languages across 61 locales, adding Odia and Urdu with instant voice cloning",
        "Backwards compatible with Sonic 3.5: existing voice clones keep their ids",
        "A dated snapshot, sonic-3.6-2026-08-27, that Cartesia says never changes"
      ],
      "cautions": [
        "Credit-based plans only (1 credit per character); the recorded rate is the Pro plan overage of $65 per 1M characters, falling to $38 on the Scale plan",
        "Cartesia does not state latency figures for this version"
      ],
      "bestFor": [
        "Multilingual speech for voice agents",
        "Indian-language and Hinglish speech"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "cartesia-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "cartesia-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "cartesia-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "cartesia-pricing"
        },
        {
          "unit": "character",
          "usd": 0.000065,
          "tier": "pro-plan-overage",
          "sourceId": "cartesia-pricing"
        },
        {
          "unit": "character",
          "usd": 0.000038,
          "tier": "scale-plan-overage",
          "sourceId": "cartesia-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "cartesia-sonic-models",
        "cartesia-pricing"
      ]
    },
    {
      "id": "gemini-3-5-transcribe",
      "providerId": "google",
      "name": "Gemini 3.5 Transcribe",
      "family": "Gemini 3.5",
      "releaseDate": "2026-08-26",
      "status": "active",
      "access": "api",
      "apiModelId": "gemini-3.5-transcribe",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "audio",
        "text"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "speech-to-text",
        "diarization"
      ],
      "inputModalities": [
        "audio"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Detects 85+ languages and handles code-switching within one recording",
        "Diarization for up to 8 speakers, word timestamps and up to 1,000 custom vocabulary terms",
        "Google estimates the blended cost at about $0.005 per minute of audio"
      ],
      "cautions": [
        "One hour of audio per request, or 30 minutes with diarization or timestamps",
        "Custom vocabulary cannot be combined with diarization or timestamps, and there is no Batch API"
      ],
      "bestFor": [
        "Batch transcription with speaker labels",
        "Multilingual meeting and call transcripts"
      ],
      "pricing": {
        "inputPerMillionUsd": 2,
        "outputPerMillionUsd": 12,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 12,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing"
        },
        {
          "unit": "audio-minute",
          "usd": 0.005,
          "tier": "estimated",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "google-gemini-3-5-transcribe-model",
        "google-gemini-3-5-changelog",
        "google-pricing"
      ]
    },
    {
      "id": "grok-imagine-video-1-5",
      "providerId": "xai",
      "name": "Grok Imagine Video 1.5",
      "family": "Grok Imagine",
      "releaseDate": null,
      "status": "active",
      "access": "api",
      "apiModelId": "grok-imagine-video-1.5",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "video"
      ],
      "taskFamilies": [
        "video"
      ],
      "capabilities": [
        "text-to-video",
        "image-to-video"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "video"
      ],
      "strengths": [
        "Native 1080p text-to-video and image-to-video, with clips from 1 to 15 seconds",
        "Videos include an audio track by default, and reference-to-video accepts voice references for up to 3 speakers",
        "Reference images and pinned first, key or last frames"
      ],
      "cautions": [
        "Editing and extending videos still use the older grok-imagine-video model",
        "xAI lists one $0.080 per second price with no breakdown by resolution",
        "xAI has not published a release date"
      ],
      "bestFor": [
        "Short 1080p clips with sound"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "xai-grok-imagine-video-1-5-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "xai-grok-imagine-video-1-5-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "xai-grok-imagine-video-1-5-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "xai-grok-imagine-video-1-5-model"
        },
        {
          "unit": "video-second",
          "usd": 0.08,
          "tier": "standard",
          "sourceId": "xai-grok-imagine-video-1-5-model"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-video",
          "benchmarkSnapshotId": "arena-text-to-video-snapshot-2026-11-05",
          "score": 1486,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-05",
          "sourceRow": "Grok Imagine Video 1.5"
        },
        {
          "benchmarkId": "arena-image-to-video",
          "benchmarkSnapshotId": "arena-image-to-video-snapshot-2026-11-05",
          "score": 1454,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-05",
          "sourceRow": "Grok Imagine Video 1.5 (720p)"
        }
      ],
      "officialSourceIds": [
        "xai-grok-imagine-video-1-5-model",
        "xai-developer-models"
      ]
    },
    {
      "id": "gemini-omni-flash-1-1",
      "providerId": "google",
      "name": "Gemini Omni Flash 1.1",
      "family": "Gemini Omni",
      "releaseDate": "2026-08-27",
      "status": "legacy",
      "access": "api",
      "apiModelId": "gemini-omni-1.1-flash",
      "contextWindowTokens": 131072,
      "maxOutputTokens": 65536,
      "modalities": [
        "text",
        "image",
        "video",
        "audio"
      ],
      "taskFamilies": [
        "video"
      ],
      "capabilities": [
        "text-to-video",
        "image-to-video",
        "video-editing"
      ],
      "inputModalities": [
        "text",
        "image",
        "video"
      ],
      "outputModalities": [
        "video",
        "audio"
      ],
      "strengths": [
        "The GA version of Google's conversational video model: clips can be refined over several turns through the Interactions API",
        "Adds video extension, first-and-last-frame interpolation and 360p to 4K output (1080p and 4K by upscaling)",
        "Video comes with sound, at about $0.10 per second for 720p"
      ],
      "cautions": [
        "Google now calls it its previous-generation video model: from December 3, 2026, requests to gemini-omni-1.1-flash are routed to Gemini Omni 1.2 Flash and billed at its rates",
        "Each clip is 3 to 10 seconds; extensions add 10 seconds at a time up to 40 seconds",
        "Editing uploaded videos is not available in every region, and voice editing is not supported"
      ],
      "bestFor": [
        "Short clips with sound that need conversational edits"
      ],
      "pricing": {
        "inputPerMillionUsd": 1.5,
        "outputPerMillionUsd": 9,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 1.5,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 9,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 17.5,
          "tier": "video-output",
          "sourceId": "google-pricing"
        },
        {
          "unit": "video-second",
          "usd": 0.1,
          "tier": "standard-720p",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-video",
          "benchmarkSnapshotId": "arena-text-to-video-snapshot-2026-11-05",
          "score": 1514,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-05",
          "sourceRow": "Gemini Omni 1.1 Flash"
        },
        {
          "benchmarkId": "arena-image-to-video",
          "benchmarkSnapshotId": "arena-image-to-video-snapshot-2026-11-05",
          "score": 1486,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-05",
          "sourceRow": "Gemini Omni 1.1 Flash"
        }
      ],
      "officialSourceIds": [
        "google-gemini-omni-flash-model",
        "google-omni-flash-guide",
        "google-gemini-3-5-changelog",
        "google-pricing",
        "google-gemini-changelog-2026-11-05"
      ]
    },
    {
      "id": "veo-3-1-lite-preview",
      "providerId": "google",
      "name": "Veo 3.1 Lite Preview",
      "family": "Veo",
      "releaseDate": "2026-03-31",
      "status": "active",
      "access": "api",
      "apiModelId": "veo-3.1-lite-generate-preview",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "video",
        "audio"
      ],
      "taskFamilies": [
        "video"
      ],
      "capabilities": [
        "text-to-video",
        "image-to-video"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "video",
        "audio"
      ],
      "strengths": [
        "Google's most cost-efficient video model, for rapid iteration and high-volume apps",
        "$0.05 per second at 720p with audio, half the price of Veo 3.1 Fast"
      ],
      "cautions": [
        "A preview with no 4K output and no video extension"
      ],
      "bestFor": [
        "High-volume, cost-sensitive video generation with sound"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "google-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing"
        },
        {
          "unit": "video-second",
          "usd": 0.05,
          "tier": "standard-720p",
          "sourceId": "google-pricing"
        },
        {
          "unit": "video-second",
          "usd": 0.08,
          "tier": "1080p",
          "sourceId": "google-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "google-veo-3-1-lite-model",
        "google-gemini-3-5-changelog",
        "google-pricing"
      ]
    },
    {
      "id": "wan-3-0-video",
      "providerId": "qwen",
      "name": "Wan 3.0 Video",
      "family": "Wan",
      "releaseDate": "2026-08-06",
      "status": "active",
      "access": "api",
      "apiModelId": "wan3.0-video",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "video",
        "audio"
      ],
      "taskFamilies": [
        "video"
      ],
      "capabilities": [
        "text-to-video",
        "image-to-video",
        "video-editing"
      ],
      "inputModalities": [
        "text",
        "image",
        "video",
        "audio"
      ],
      "outputModalities": [
        "video"
      ],
      "strengths": [
        "One model for text-to-video, image-to-video from a first frame or first and last frames, and reference-based generation, up to 30 seconds at 30fps",
        "A request can combine up to 10 reference images, 5 video clips and 5 audio clips",
        "$0.10 per second at 720P list price, $0.05 at 480P and $0.20 at 1080P"
      ],
      "cautions": [
        "Alibaba labels it a preview",
        "Billing counts input video seconds as well as output, so reference clips add cost",
        "The independent leaderboards list a Wan 3.0 with a later date than this API model, so no score is matched to it"
      ],
      "bestFor": [
        "Long single-shot clips",
        "Reference-driven video mixing images, video and audio"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "alibaba-model-studio-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "alibaba-model-studio-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "alibaba-model-studio-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "alibaba-model-studio-pricing"
        },
        {
          "unit": "video-second",
          "usd": 0.1,
          "tier": "standard-720p",
          "sourceId": "alibaba-model-studio-pricing"
        },
        {
          "unit": "video-second",
          "usd": 0.05,
          "tier": "480p",
          "sourceId": "alibaba-model-studio-pricing"
        },
        {
          "unit": "video-second",
          "usd": 0.2,
          "tier": "1080p",
          "sourceId": "alibaba-model-studio-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "alibaba-wan-3-video-api",
        "alibaba-model-studio-new-models",
        "alibaba-model-studio-pricing"
      ]
    },
    {
      "id": "seedance-2-5",
      "providerId": "bytedance",
      "name": "Dreamina Seedance 2.5",
      "family": "Seedance",
      "releaseDate": null,
      "status": "active",
      "access": "api",
      "apiModelId": "dreamina-seedance-2-5-260628",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "video"
      ],
      "taskFamilies": [
        "video"
      ],
      "capabilities": [
        "text-to-video",
        "image-to-video",
        "video-editing"
      ],
      "inputModalities": [
        "text",
        "image",
        "video"
      ],
      "outputModalities": [
        "video"
      ],
      "strengths": [
        "Up to 30 seconds of video per request, multi-round extension, up to 50 reference assets and timestamp-precise editing",
        "Text-to-video, image-to-video, reference-to-video, editing and extension at 480p to 1080p, 24 fps"
      ],
      "cautions": [
        "Billed by tokens, not seconds: the recorded per-second rates are BytePlus estimates for a 5-second 16:9 clip without video input",
        "Requests with video input count the input duration and carry a minimum charge",
        "BytePlus dates the release only to June 2026"
      ],
      "bestFor": [
        "Multi-shot clips up to 30 seconds",
        "Timestamp-precise video editing"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "byteplus-modelark-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "byteplus-modelark-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "byteplus-modelark-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "byteplus-modelark-pricing"
        },
        {
          "unit": "video-second",
          "usd": 0.231,
          "tier": "standard-720p",
          "sourceId": "byteplus-modelark-pricing"
        },
        {
          "unit": "video-second",
          "usd": 0.103,
          "tier": "480p",
          "sourceId": "byteplus-modelark-pricing"
        },
        {
          "unit": "video-second",
          "usd": 0.569,
          "tier": "1080p",
          "sourceId": "byteplus-modelark-pricing"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-video",
          "benchmarkSnapshotId": "arena-text-to-video-snapshot-2026-11-05",
          "score": 1480,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-05",
          "sourceRow": "Dreamina Seedance 2.5 (720p)"
        },
        {
          "benchmarkId": "arena-image-to-video",
          "benchmarkSnapshotId": "arena-image-to-video-snapshot-2026-11-05",
          "score": 1488,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-05",
          "sourceRow": "Dreamina Seedance 2.5 (720p)"
        }
      ],
      "officialSourceIds": [
        "byteplus-modelark-releases",
        "byteplus-modelark-pricing"
      ]
    },
    {
      "id": "muse-spark-open",
      "providerId": "meta",
      "name": "Muse Spark Open",
      "family": "Muse Spark",
      "releaseDate": "2026-10-15",
      "status": "active",
      "access": "open-weights",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "video",
        "audio"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "agentic-coding",
        "image-understanding",
        "video-understanding",
        "audio-understanding"
      ],
      "inputModalities": [
        "text",
        "image",
        "video",
        "audio"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "An 816B mixture-of-experts model with 42B active parameters and a 1M-token context that takes text, image, video and audio input",
        "Meta reports 89.0% on GPQA Diamond and 76.4% on Terminal-Bench 2.1 at its highest reasoning setting",
        "Weights on Hugging Face, Amazon SageMaker AI, Google Cloud Model Garden and Microsoft Foundry, with hosted access through Together AI, Fireworks, Groq and OpenRouter"
      ],
      "cautions": [
        "The Muse Model License limits commercial use; it is not an open-source licence",
        "Meta does not serve it on its own Model API, so hosted prices come from third parties",
        "Self-hosting needs at least 700GB of GPU memory in NVFP4 or 2TB in BF16"
      ],
      "bestFor": [
        "Self-hosted coding and agent work on owned infrastructure",
        "Long-context multimodal analysis where data must stay in-house"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "meta-muse-spark-open-weights"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "meta-muse-spark-open-weights"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "meta-muse-spark-open-weights"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "meta-muse-spark-open-weights"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "epoch-gpqa-diamond",
          "benchmarkSnapshotId": "epoch-gpqa-diamond-snapshot-2026-11-03",
          "score": 88.6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "Muse-Spark-Open-816B-A42B_xhigh"
        },
        {
          "benchmarkId": "epoch-frontiermath-tiers-1-3-v2",
          "benchmarkSnapshotId": "epoch-frontiermath-tiers-1-3-v2-snapshot-2026-11-03",
          "score": 69.8,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "Muse-Spark-Open-816B-A42B_xhigh"
        },
        {
          "benchmarkId": "epoch-simpleqa-verified",
          "benchmarkSnapshotId": "epoch-simpleqa-verified-snapshot-2026-11-03",
          "score": 56.6,
          "sourceId": "epoch-benchmark-data-2026-11-03",
          "measuredAt": "2026-11-03",
          "sourceRow": "Muse-Spark-Open-816B-A42B_xhigh"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1526,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Muse Spark Open (Max)"
        }
      ],
      "officialSourceIds": [
        "meta-muse-spark-open-release",
        "meta-muse-spark-open-weights"
      ]
    },
    {
      "id": "nemotron-3-5-super-vl",
      "providerId": "nvidia",
      "name": "Nemotron 3.5 Super VL",
      "family": "Nemotron 3.5",
      "releaseDate": "2026-10-15",
      "status": "active",
      "access": "open-weights",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "agentic-coding",
        "image-understanding",
        "document-understanding",
        "chart-understanding",
        "ocr",
        "video-understanding"
      ],
      "inputModalities": [
        "text",
        "image",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "A 120B mixture-of-experts vision-language model with 12B active parameters and up to 1M tokens of context that reads images, charts, documents and videos of up to two minutes",
        "NVIDIA reports 89.5% on GPQA Diamond and 71.1% on SWE-bench Verified in BF16, with the NVFP4 build within a point on most tests",
        "Open weights, data and training recipes under the OpenMDW-1.1 licence, on Hugging Face, OpenRouter and ModelScope and as an NVIDIA NIM"
      ],
      "cautions": [
        "NVIDIA calls it 5x faster than the next best open model in its class; that speed claim is not measured independently here",
        "NVIDIA sells no per-token API for it, so hosted prices come from third parties"
      ],
      "bestFor": [
        "Agents that read screenshots, charts and documents",
        "Self-hosted multimodal agents on NVIDIA hardware"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "nvidia-nemotron-3-5-super-vl-weights"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "nvidia-nemotron-3-5-super-vl-weights"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "nvidia-nemotron-3-5-super-vl-weights"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "nvidia-nemotron-3-5-super-vl-weights"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1412,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "NVIDIA-Nemotron-3.5-Super-VL-120B-A12B (NVFP4)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1399,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "NVIDIA-Nemotron-3.5-Super-VL-120B-A12B (NVFP4)"
        }
      ],
      "officialSourceIds": [
        "nvidia-nemotron-3-5-super-vl-release",
        "nvidia-nemotron-3-5-super-vl-weights"
      ]
    },
    {
      "id": "nemotron-3-5-bolt-74b",
      "providerId": "nvidia",
      "name": "Nemotron 3.5 Bolt 74B-A9B",
      "family": "Nemotron 3.5",
      "releaseDate": "2026-10-15",
      "status": "active",
      "access": "open-weights",
      "contextWindowTokens": 262144,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "agentic-coding",
        "image-understanding",
        "video-understanding"
      ],
      "inputModalities": [
        "text",
        "image",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "A 74B mixture-of-experts model with 9B active parameters, compressed from Nemotron 3.5 Super VL and natively quantized to 2 bits to fit in 24GB of memory",
        "NVIDIA reports 86.4% on GPQA Diamond and 51.0% on Terminal-Bench 2.1 for the 24GB build, ahead of Qwen3.6-35B-A3B on its own table",
        "Runs in llama.cpp, Ollama, LM Studio and vLLM on RTX PCs, RTX Spark laptops and DGX Spark, with up to 256K tokens of context"
      ],
      "cautions": [
        "Usable context depends on memory: the recipes in its model card use 64K tokens at 28GB and 256K at 32GB",
        "NVIDIA sells no per-token API for it, and no independent benchmark score is recorded here yet"
      ],
      "bestFor": [
        "Local agents on a single 24GB to 32GB GPU or laptop",
        "Private multimodal assistants that must run offline"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "nvidia-nemotron-3-5-bolt-weights"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "nvidia-nemotron-3-5-bolt-weights"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "nvidia-nemotron-3-5-bolt-weights"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "nvidia-nemotron-3-5-bolt-weights"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "nvidia-nemotron-3-5-bolt-weights",
        "nvidia-nemotron-3-5-super-vl-release"
      ]
    },
    {
      "id": "beam-1-0",
      "providerId": "reflection",
      "name": "Beam 1.0",
      "family": "Beam",
      "releaseDate": "2026-10-16",
      "status": "active",
      "access": "both",
      "apiModelId": "Beam-1.0-501B-A23B",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning",
        "coding",
        "agentic-coding"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "A 501B sparse mixture-of-experts model with 23B active parameters and a 1M-token context, built for coding, reasoning and agents",
        "Apache 2.0 weights in BF16, FP8 and NVFP4, with no ban on distillation, fine-tuning or training on its outputs",
        "Reflection reports 81.6% on SWE-bench Verified and 90.6% on GPQA Diamond for the BF16 weights"
      ],
      "cautions": [
        "Reflection API is a public beta that is free to start, with no published per-token price",
        "Text only, with no image, audio or video input"
      ],
      "bestFor": [
        "Self-hosted coding agents under a permissive licence",
        "Fine-tuning or distilling a large open model"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "reflection-developer-docs"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "reflection-developer-docs"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "reflection-developer-docs"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "reflection-developer-docs"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1436,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Beam 1.0 501B A23B"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1436,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Beam 1.0 501B A23B"
        }
      ],
      "officialSourceIds": [
        "reflection-beam-release",
        "reflection-beam-weights",
        "reflection-developer-docs"
      ]
    },
    {
      "id": "minimax-m3-1-flash",
      "providerId": "minimax",
      "name": "MiniMax M3.1 Flash",
      "family": "MiniMax M3",
      "releaseDate": "2026-10-19",
      "status": "active",
      "access": "both",
      "apiModelId": "MiniMax-M3.1-Flash",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "agentic-coding",
        "image-understanding",
        "video-understanding"
      ],
      "inputModalities": [
        "text",
        "image",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "A mixture-of-experts model with 23B active parameters and a 1M-token context that reads text, images and video",
        "MiniMax reports 73.7% on DeepSWE v1.1 and nearly three times the online inference speed of its predecessor at about 70% lower cost on agent tasks",
        "Low list price of $0.15 input and $0.60 output per 1M tokens, with weights on Hugging Face under the MiniMax community licence"
      ],
      "cautions": [
        "Thinking cannot be turned off; effort runs from low to max, and max is the default",
        "MiniMax gives two parameter counts (about 428B in the model card, 220B in the file metadata)"
      ],
      "bestFor": [
        "Low-cost coding and agent loops",
        "Long-context multimodal work on a budget"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.15,
        "outputPerMillionUsd": 0.6,
        "cachedInputPerMillionUsd": 0.015,
        "sourceId": "minimax-m3-1-flash-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.15,
          "tier": "standard",
          "sourceId": "minimax-m3-1-flash-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": 0.6,
          "tier": "standard",
          "sourceId": "minimax-m3-1-flash-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.015,
          "tier": "cached",
          "sourceId": "minimax-m3-1-flash-release"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1582,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "MiniMax-M3.1-Flash"
        }
      ],
      "officialSourceIds": [
        "minimax-m3-1-flash-release",
        "minimax-m3-1-flash-weights",
        "minimax-text-generation-docs"
      ]
    },
    {
      "id": "mercury-3",
      "providerId": "inception",
      "name": "Mercury 3",
      "family": "Mercury",
      "releaseDate": "2026-10-19",
      "status": "active",
      "access": "api",
      "apiModelId": "mercury-3",
      "contextWindowTokens": 1000000,
      "maxOutputTokens": 131072,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "A diffusion language model that writes many tokens in parallel; Inception reports 1,068 tokens per second on NVIDIA GPUs",
        "A 1M-token context, up from 260K on Mercury 2.5, with up to 131,072 output tokens and reasoning efforts from instant to high",
        "Inception says it outperforms GPT-6 Luna and Claude Haiku 5.5 across its benchmark set while running 4 to 8 times faster"
      ],
      "cautions": [
        "The recorded rates are list prices; a 50% launch discount was running with no published end date",
        "Inception comparisons are its own, and no independent benchmark score is recorded here yet"
      ],
      "bestFor": [
        "Latency-critical agent loops and real-time search agents",
        "High-volume text generation at low cost"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.2,
        "outputPerMillionUsd": 0.7,
        "cachedInputPerMillionUsd": 0.02,
        "sourceId": "inception-models"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.2,
          "tier": "standard",
          "sourceId": "inception-models"
        },
        {
          "unit": "million-output-tokens",
          "usd": 0.7,
          "tier": "standard",
          "sourceId": "inception-models"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.02,
          "tier": "cached",
          "sourceId": "inception-models"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "inception-models",
        "inception-mercury-3-release"
      ]
    },
    {
      "id": "grok-voice-3",
      "providerId": "xai",
      "name": "Grok Voice 3.0",
      "family": "Grok Voice",
      "releaseDate": "2026-10-14",
      "status": "active",
      "access": "api",
      "apiModelId": "grok-voice-3.0",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "audio-understanding",
        "realtime-voice"
      ],
      "inputModalities": [
        "text",
        "audio"
      ],
      "outputModalities": [
        "text",
        "audio"
      ],
      "strengths": [
        "A speech-to-speech model for real-time voice agents with tool use, on the same realtime endpoint as Grok Voice 2.0 with an OpenAI-compatible migration path",
        "xAI cites Artificial Analysis results of 86.9 on its Speech-to-Speech Index (Grok Voice 2.0: 81.3) and 98.5% on Big Bench Audio",
        "Priced at $0.08 per minute, the same as Grok Voice 2.0, which the grok-voice-latest alias no longer points to"
      ],
      "cautions": [
        "The per-model page leaves pricing blank; the rate comes from the launch post",
        "The benchmark figures are cited by xAI and are not recorded here as independent scores"
      ],
      "bestFor": [
        "Customer-support and in-car voice agents",
        "Real-time voice assistants that call tools"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "xai-grok-voice-3-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "xai-grok-voice-3-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "xai-grok-voice-3-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "xai-grok-voice-3-release"
        },
        {
          "unit": "realtime-minute",
          "usd": 0.08,
          "tier": "standard",
          "sourceId": "xai-grok-voice-3-release"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "xai-grok-voice-3-release",
        "xai-grok-voice-3-model"
      ]
    },
    {
      "id": "voxtral-small-transcribe-3",
      "providerId": "mistral",
      "name": "Voxtral Small Transcribe 3",
      "family": "Voxtral",
      "releaseDate": "2026-10-20",
      "status": "active",
      "access": "api",
      "apiModelId": "voxtral-small-transcribe-3-0",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "audio",
        "text"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "speech-to-text",
        "diarization",
        "speech-translation",
        "audio-understanding"
      ],
      "inputModalities": [
        "audio"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "The flagship Mistral transcription and diarization model, with timestamps, context biasing and more than 50 languages",
        "Also translates speech and answers questions about audio through the same API",
        "Priced at $0.004 per minute; Mistral says it outperforms Whisper, Scribe v2 and Gemini 3.5 Transcribe on its own multilingual tests"
      ],
      "cautions": [
        "Released as a public preview, so behaviour and pricing may change",
        "Its comparisons with other transcription models are Mistral tests"
      ],
      "bestFor": [
        "Multilingual meeting and call transcription with speaker labels",
        "Speech pipelines that must stay on EU-hosted infrastructure"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "mistral-voxtral-small-transcribe-3-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "mistral-voxtral-small-transcribe-3-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "mistral-voxtral-small-transcribe-3-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "mistral-voxtral-small-transcribe-3-model"
        },
        {
          "unit": "audio-minute",
          "usd": 0.004,
          "tier": "standard",
          "sourceId": "mistral-voxtral-small-transcribe-3-model"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "mistral-voxtral-3-release",
        "mistral-voxtral-small-transcribe-3-model"
      ]
    },
    {
      "id": "voxtral-mini-tts-3",
      "providerId": "mistral",
      "name": "Voxtral Mini TTS 3",
      "family": "Voxtral",
      "releaseDate": "2026-10-20",
      "status": "active",
      "access": "api",
      "apiModelId": "voxtral-mini-tts-3-0",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "text-to-speech"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "audio"
      ],
      "strengths": [
        "Text-to-speech for batch and real-time use, with zero-shot and professional voice cloning",
        "Mistral reports about 90 ms to the first audio when streaming",
        "Priced at $13 per 1M characters"
      ],
      "cautions": [
        "Language coverage is unclear: the launch post says 35+ languages and 50+ dialects, while the model page lists nine",
        "Released as a public preview, so behaviour and pricing may change"
      ],
      "bestFor": [
        "Multilingual voice agents and narration",
        "Branded voices cloned from recorded samples"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "mistral-voxtral-mini-tts-3-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "mistral-voxtral-mini-tts-3-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "mistral-voxtral-mini-tts-3-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "mistral-voxtral-mini-tts-3-model"
        },
        {
          "unit": "character",
          "usd": 0.000013,
          "tier": "standard",
          "sourceId": "mistral-voxtral-mini-tts-3-model"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "mistral-voxtral-3-release",
        "mistral-voxtral-mini-tts-3-model"
      ]
    },
    {
      "id": "elorian-glimpse",
      "providerId": "elorian",
      "name": "Glimpse",
      "family": "Glimpse",
      "releaseDate": "2026-10-16",
      "status": "active",
      "access": "api",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "image-understanding",
        "video-understanding",
        "chart-understanding",
        "temporal-reasoning"
      ],
      "inputModalities": [
        "text",
        "image",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "A visual thinking model that reasons about the physical world from images and video, released as a research preview through Elorian's API and chat",
        "Elorian reports 92.90 on CharXiv-R chart reasoning, 82.66 on UGC-VideoCap captioning and 83.36 on FSC-147 counting, ahead of GPT-6 Astra on all three",
        "$0.60 input and $3 output per 1M tokens, which Elorian puts 94% below GPT-6 Astra"
      ],
      "cautions": [
        "In Elorian's own table GPT-6 Astra scores higher on six of the twelve evaluations, including TOMATO temporal reasoning and Video-MME v2",
        "Post-trained on an open-weights base model Elorian does not name; it publishes no context window or output limit, and no independent benchmark score is recorded here yet"
      ],
      "bestFor": [
        "Visual question answering over charts, scenes and short videos",
        "Counting and captioning at a lower price than frontier multimodal models"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.6,
        "outputPerMillionUsd": 3,
        "cachedInputPerMillionUsd": null,
        "sourceId": "elorian-glimpse-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.6,
          "tier": "standard",
          "sourceId": "elorian-glimpse-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": 3,
          "tier": "standard",
          "sourceId": "elorian-glimpse-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "elorian-glimpse-release"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "elorian-glimpse-release",
        "elorian-models"
      ]
    },
    {
      "id": "mai-voice-2-1",
      "providerId": "microsoft",
      "name": "MAI-Voice-2.1",
      "family": "MAI-Voice",
      "releaseDate": "2026-10-01",
      "status": "active",
      "access": "api",
      "apiModelId": "MAI-Voice-2.1",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "text-to-speech"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "audio"
      ],
      "strengths": [
        "23 languages and 26 locales, and one voice speaks every supported language with a native accent",
        "Clones a voice from a 5 to 60 second sample and controls tone and delivery per turn or sentence, SSML speaking styles included",
        "Long-form narration and multi-speaker dialogue"
      ],
      "cautions": [
        "Microsoft Foundry lists it as a preview, and voice cloning needs Microsoft approval",
        "No independent benchmark score is recorded here yet"
      ],
      "bestFor": [
        "Multilingual products that keep one voice across languages",
        "Narration and dialogue with controlled delivery"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "microsoft-mai-streaming-voice-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "microsoft-mai-streaming-voice-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "microsoft-mai-streaming-voice-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "microsoft-mai-streaming-voice-release"
        },
        {
          "unit": "character",
          "usd": 0.000022,
          "tier": "standard",
          "sourceId": "microsoft-mai-streaming-voice-release"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "microsoft-mai-streaming-voice-release",
        "microsoft-foundry-mai-voice-2-1"
      ]
    },
    {
      "id": "mai-voice-2-1-flash",
      "providerId": "microsoft",
      "name": "MAI-Voice-2.1-Flash",
      "family": "MAI-Voice",
      "releaseDate": "2026-10-01",
      "status": "active",
      "access": "api",
      "apiModelId": "MAI-Voice-2.1-Flash",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "text-to-speech"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "audio"
      ],
      "strengths": [
        "Microsoft reports 45 seconds of audio generated with 150 ms end-to-end latency",
        "The same languages and cross-language voices as MAI-Voice-2.1, with streaming output and barge-in handling for voice agents",
        "$15 per 1M characters, about 32% below MAI-Voice-2.1"
      ],
      "cautions": [
        "Microsoft Foundry lists it as a preview, and voice cloning needs Microsoft approval",
        "Microsoft's claims of 55% faster inference and about 60% lower cost do not name the models compared"
      ],
      "bestFor": [
        "Voice agents and call-center assistants",
        "High-volume, low-latency speech"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "microsoft-mai-streaming-voice-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "microsoft-mai-streaming-voice-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "microsoft-mai-streaming-voice-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "microsoft-mai-streaming-voice-release"
        },
        {
          "unit": "character",
          "usd": 0.000015,
          "tier": "standard",
          "sourceId": "microsoft-mai-streaming-voice-release"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "microsoft-mai-streaming-voice-release",
        "microsoft-foundry-mai-voice-2-1-flash"
      ]
    },
    {
      "id": "mai-transcribe-2-streaming",
      "providerId": "microsoft",
      "name": "MAI-Transcribe-2-Streaming",
      "family": "MAI-Transcribe",
      "releaseDate": "2026-10-01",
      "status": "active",
      "access": "api",
      "apiModelId": "MAI-Transcribe-2-Streaming",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "audio",
        "text"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "speech-to-text"
      ],
      "inputModalities": [
        "audio"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Microsoft reports first partial transcripts just over 100 ms after audio arrives, and first place for accuracy on final and partial transcripts at Artificial Analysis",
        "60 languages with automatic, continuous language detection",
        "$0.54 per hour of audio at launch"
      ],
      "cautions": [
        "Microsoft Foundry lists it as a preview",
        "The $0.54 per hour price is introductory until the end of 2026, and the later price is not published"
      ],
      "bestFor": [
        "Live captions and voice agents",
        "Multilingual calls where speakers switch languages"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "microsoft-mai-streaming-voice-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "microsoft-mai-streaming-voice-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "microsoft-mai-streaming-voice-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "microsoft-mai-streaming-voice-release"
        },
        {
          "unit": "audio-minute",
          "usd": 0.009,
          "tier": "standard",
          "effectiveUntil": "2026-12-31",
          "sourceId": "microsoft-mai-streaming-voice-release"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "microsoft-mai-streaming-voice-release",
        "microsoft-foundry-mai-transcribe-2-streaming"
      ]
    },
    {
      "id": "longcat-2-5",
      "providerId": "meituan",
      "name": "LongCat-2.5",
      "family": "LongCat-2.5",
      "releaseDate": "2026-11-02",
      "status": "active",
      "access": "api",
      "apiModelId": "LongCat-2.5",
      "contextWindowTokens": 1048576,
      "maxOutputTokens": 262144,
      "modalities": [
        "text",
        "image",
        "video"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "agentic-coding",
        "image-understanding",
        "video-understanding"
      ],
      "inputModalities": [
        "text",
        "image",
        "video"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Meituan built it for long-horizon agents that pair visual understanding with coding: it reads images, and videos by URL, alongside text",
        "A 1,048,576-token context, up to 262,144 output tokens and a thinking budget of up to 262,144 tokens",
        "OpenAI- and Anthropic-compatible APIs, with setup guides for Claude Code, Codex, OpenCode and other coding agents"
      ],
      "cautions": [
        "The recorded rates are a limited-time discount: Meituan publishes no list price and no end date",
        "API only: no weights had been published when this record was made",
        "Released on 2 November 2026, so no independent benchmark score is recorded here yet"
      ],
      "bestFor": [
        "Coding agents working through whole projects in Claude Code, Codex or similar harnesses",
        "Visual coding and design work that mixes screenshots or screen recordings with code"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.3,
        "outputPerMillionUsd": 1.2,
        "cachedInputPerMillionUsd": 0.006,
        "sourceId": "meituan-longcat-2-5-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.3,
          "tier": "standard",
          "sourceId": "meituan-longcat-2-5-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 1.2,
          "tier": "standard",
          "sourceId": "meituan-longcat-2-5-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.006,
          "tier": "cached",
          "sourceId": "meituan-longcat-2-5-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "meituan-longcat-changelog",
        "meituan-longcat-model-params",
        "meituan-longcat-visual-understanding",
        "meituan-longcat-2-5-pricing"
      ]
    },
    {
      "id": "assemblyai-tts-1",
      "providerId": "assemblyai",
      "name": "AssemblyAI-TTS-1",
      "family": "AssemblyAI-TTS",
      "releaseDate": "2026-10-28",
      "status": "active",
      "access": "api",
      "apiModelId": "assemblyai-tts-1",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "audio"
      ],
      "taskFamilies": [
        "audio"
      ],
      "capabilities": [
        "text-to-speech"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "audio"
      ],
      "strengths": [
        "$15 per 1M characters with every voice, all six languages, instant voice cloning, word timestamps and US or EU processing included, and no concurrency fees or minimums",
        "Built for voice agents: AssemblyAI quotes 90 ms to first audio in production, streamed over WebSocket, HTTP or an OpenAI-compatible route",
        "More than 4,000 voices, each speaking English, Spanish, French, German, Italian and Portuguese"
      ],
      "cautions": [
        "Six languages at launch; AssemblyAI says more languages and accents will follow",
        "No fine-tuned (Pro) voice cloning, pronunciation dictionaries or emotion tags at launch",
        "The head-to-head figures in the launch post are AssemblyAI tests"
      ],
      "bestFor": [
        "Voice agents on live phone calls",
        "High-volume speech where the cost per character matters"
      ],
      "pricing": {
        "inputPerMillionUsd": null,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": null,
        "sourceId": "assemblyai-models"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "assemblyai-models"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "assemblyai-models"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "assemblyai-models"
        },
        {
          "unit": "character",
          "usd": 0.000015,
          "tier": "standard",
          "sourceId": "assemblyai-models"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "assemblyai-tts-1-release",
        "assemblyai-models"
      ]
    },
    {
      "id": "hy4",
      "providerId": "tencent",
      "name": "Hy4",
      "family": "Hy4",
      "releaseDate": "2026-11-04",
      "status": "active",
      "access": "api",
      "apiModelId": "hy4",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "agentic-coding",
        "image-understanding",
        "chart-understanding"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Tencent's flagship and the successor to the open Hy4 preview (a 770B-parameter mixture-of-experts, 49B active per token): it adds image understanding and, for the first time, computer use across command-line and graphical apps",
        "In Tencent's runs it scores 74.9% on DeepSWE v1.1 and 75.2% on APEX-Agents v1.1, against 72.6% and 74.8% for Claude Opus 5.5",
        "$0.834 input, $2.501 output and $0.042 cached input per 1M tokens, the same as Hy4 preview"
      ],
      "cautions": [
        "API only for now: Tencent says it will publish the weights and a system card later, and only Hy4 preview (August 2026, Apache 2.0) is open today",
        "In Tencent's own table it scores 44.4% on Terminal-Bench 4.0 against 66.8% for Kimi K3.1 Pro, and Tencent says a clear gap remains to top proprietary models on research-level math, long-horizon coding and hard visual questions",
        "Tencent publishes no context window or output limit for Hy4 (the preview card lists 1M tokens), and no independent benchmark score is recorded here yet"
      ],
      "bestFor": [
        "Agents that work across terminals, desktop apps and browsers on a tight budget",
        "Finance and office work: research reports, spreadsheets and slides"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.834,
        "outputPerMillionUsd": 2.501,
        "cachedInputPerMillionUsd": 0.042,
        "sourceId": "tencent-hy4-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.834,
          "tier": "standard",
          "sourceId": "tencent-hy4-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": 2.501,
          "tier": "standard",
          "sourceId": "tencent-hy4-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.042,
          "tier": "cached",
          "sourceId": "tencent-hy4-release"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "tencent-hy4-release",
        "tencent-hy4-preview-card"
      ]
    },
    {
      "id": "gemini-4-carbon",
      "providerId": "google",
      "name": "Gemini 4 Carbon",
      "family": "Gemini 4",
      "releaseDate": "2026-10-28",
      "status": "active",
      "access": "limited",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "coding",
        "agentic-coding",
        "image-understanding",
        "chart-understanding"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Google's strongest model at launch, built for cyber defense and long-horizon engineering: Google reports 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation and 71.0% on CWE-bench for patching, each ahead of GPT-6 Astra in its table",
        "Arena rates it 1528 on Arena Text and 1768 on Arena WebDev (October 28, 2026), above Gemini 4 Argon on both",
        "Google reports 96.2% on GraphWalks from 256K to 1M tokens, against 84.2% for Gemini 4 Argon, and an introductory price of $2 input and $10 output per 1M tokens, the same as Argon"
      ],
      "cautions": [
        "Limited access: Google gives it first to a small group of cyber defenders, then to more trusted partners through its Fairwind Program, which governments, healthcare providers and telecoms apply to; there is no public API or app access yet",
        "The $2 and $10 rates are introductory, with cached input 95% off; after an introductory period Google has not dated, $4 input and $20 output per 1M tokens apply",
        "In Google's own table it trails GPT-6 Astra on FrontierSWE 2 (61.2% against 65.5%) and Claude Opus 5.5 on OSWorld-2.1 (81.5% against 87.2%)",
        "Google publishes no context window or output limit for Carbon yet"
      ],
      "bestFor": [
        "Finding, validating and patching vulnerabilities, for teams in Google's Fairwind Program",
        "Long-horizon engineering over very long inputs, once access opens"
      ],
      "pricing": {
        "inputPerMillionUsd": 2,
        "outputPerMillionUsd": 10,
        "cachedInputPerMillionUsd": 0.1,
        "sourceId": "google-gemini-4-carbon-release"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "standard",
          "sourceId": "google-gemini-4-carbon-release"
        },
        {
          "unit": "million-output-tokens",
          "usd": 10,
          "tier": "standard",
          "sourceId": "google-gemini-4-carbon-release"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.1,
          "tier": "cached",
          "sourceId": "google-gemini-4-carbon-release"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text",
          "benchmarkSnapshotId": "arena-text-snapshot-2026-10-28",
          "score": 1528,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 4 Carbon (High)"
        },
        {
          "benchmarkId": "arena-webdev",
          "benchmarkSnapshotId": "arena-webdev-snapshot-2026-10-28",
          "score": 1768,
          "sourceId": "arena-leaderboard-dataset-2026-11-03",
          "measuredAt": "2026-10-28",
          "sourceRow": "Gemini 4 Carbon (High)"
        }
      ],
      "officialSourceIds": [
        "google-gemini-4-carbon-release",
        "google-fairwind-program"
      ]
    },
    {
      "id": "gpt-rosalind-6",
      "providerId": "openai",
      "name": "GPT-Rosalind-6",
      "family": "GPT-Rosalind",
      "releaseDate": "2026-10-30",
      "status": "active",
      "access": "limited",
      "apiModelId": "gpt-6-rosalind",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text"
      ],
      "taskFamilies": [
        "text-nlp"
      ],
      "capabilities": [
        "reasoning",
        "knowledge",
        "data-analysis"
      ],
      "inputModalities": [
        "text"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "OpenAI's reasoning model for biology, drug discovery and translational medicine, on the same weights as GPT-6 Astra but with fewer refusals on dual-use biology questions",
        "OpenAI reports 49.6% on MedChemBench v7, 37.7% on GeneBench Pro v13 and 61.3% on LifeSciBench Gold at its highest reasoning budget, up from 28.9%, 16.7% and 44.0% for GPT-Rosalind-5.5",
        "$15 input, $75 output and $1.50 cached input per 1M tokens, 1.5 times the price of GPT-6 Astra"
      ],
      "cautions": [
        "Limited access: OpenAI offers it only through its trusted-access program, after reviewing each applicant's intended use, security controls and governance, for internal research",
        "OpenAI treats it as High capability in the biological and chemical domain, so access comes with security, monitoring and enforcement terms",
        "OpenAI publishes no context window for it and compares it only with GPT-Rosalind-5.5; no independent benchmark score is recorded here yet"
      ],
      "bestFor": [
        "Drug discovery and life sciences research teams approved for OpenAI's trusted access",
        "Dual-use biology research that GPT-6 Astra is more likely to refuse"
      ],
      "pricing": {
        "inputPerMillionUsd": 15,
        "outputPerMillionUsd": 75,
        "cachedInputPerMillionUsd": 1.5,
        "sourceId": "openai-pricing-life-sciences"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 15,
          "tier": "standard",
          "sourceId": "openai-pricing-life-sciences"
        },
        {
          "unit": "million-output-tokens",
          "usd": 75,
          "tier": "standard",
          "sourceId": "openai-pricing-life-sciences"
        },
        {
          "unit": "million-input-tokens",
          "usd": 1.5,
          "tier": "cached",
          "sourceId": "openai-pricing-life-sciences"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "openai-gpt-rosalind-6-system-card",
        "openai-pricing-life-sciences"
      ]
    },
    {
      "id": "gpt-6-cyber",
      "providerId": "openai",
      "name": "GPT-6 Cyber",
      "family": "GPT Cyber",
      "releaseDate": "2026-11-05",
      "status": "active",
      "access": "limited",
      "apiModelId": "gpt-6-cyber",
      "contextWindowTokens": 1050000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "OpenAI's cybersecurity model for authorized vulnerability research and exploit development: a variant of GPT-6 Astra trained to find and validate exploitable vulnerabilities while staying within the scope a researcher sets",
        "OpenAI reports 48.5% on ExploitGym, against 32.1% for GPT-5.6 Cyber and 42.4% for GPT-6 Astra, and 92.8% on SRE-Bench reverse engineering against 88.0% for Astra",
        "OpenAI says it built working exploits for 948 Linux kernel vulnerabilities, more than five times the 166 GPT-5.6 Sol managed with the same harness, and found 81 previously unknown high-severity Firefox vulnerabilities, many already fixed by Mozilla",
        "1,050,000-token context window and 128,000 max output tokens, at $20 input, $100 output and $2 cached input per 1M tokens, twice the standard rates of GPT-6 Astra"
      ],
      "cautions": [
        "Limited access: OpenAI is rolling it out through the API and Codex to select approved Daybreak Red customers first, with identity verification, single sign-on, multi-factor authentication and API IP allowlisting",
        "OpenAI rates it Critical in cybersecurity under its Preparedness Framework, like GPT-6 Astra and up from High for GPT-5.6 Cyber, and adds checks that block high-risk activity against critical systems",
        "It does not beat GPT-6 Astra at general coding (82.95% against 83.24% on OpenAI's internal ExpertSWE), and on Irregular's FrontierCyber it solved 90 of 288 challenges against 96 for Astra; no independent benchmark score is recorded here yet"
      ],
      "bestFor": [
        "Security teams approved for Daybreak Red that develop and validate exploits in authorized tests",
        "Vulnerability research and patch validation across large codebases such as browsers and kernels"
      ],
      "pricing": {
        "inputPerMillionUsd": 20,
        "outputPerMillionUsd": 100,
        "cachedInputPerMillionUsd": 2,
        "sourceId": "openai-gpt-6-cyber-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 20,
          "tier": "standard",
          "sourceId": "openai-gpt-6-cyber-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": 100,
          "tier": "standard",
          "sourceId": "openai-gpt-6-cyber-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 2,
          "tier": "cached",
          "sourceId": "openai-gpt-6-cyber-model"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "openai-gpt-6-cyber-release",
        "openai-gpt-6-cyber-system-card",
        "openai-gpt-6-cyber-model"
      ]
    },
    {
      "id": "gpt-56-cyber",
      "providerId": "openai",
      "name": "GPT-5.6 Cyber",
      "family": "GPT Cyber",
      "releaseDate": "2026-08-10",
      "status": "active",
      "access": "limited",
      "apiModelId": "gpt-5.6-cyber",
      "contextWindowTokens": 400000,
      "maxOutputTokens": 128000,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "OpenAI's cybersecurity model for Daybreak Red before GPT-6 Cyber, built on GPT-5.6 Sol and trained to find zero-day vulnerabilities, build exploit chains and refuse fewer dual-use security tasks",
        "OpenAI reports it completes 95.0% of advanced cybersecurity requests in its internal evaluation, against 1.5% for GPT-5.6 Sol and 57.3% for GPT-5.5 Cyber, and says it found two chained V8 vulnerabilities and over 400 kernel privilege-escalation flaws",
        "400,000-token context window and 128,000 max output tokens, at $12.50 input, $75 output and $1.25 cached input per 1M tokens"
      ],
      "cautions": [
        "Limited access: available only to approved Daybreak Red customers",
        "OpenAI rates it High but not Critical in cybersecurity, and says it sometimes writes shorter, less detailed vulnerability reports than GPT-5.6 Sol",
        "OpenAI's newer GPT-6 Cyber (November 5, 2026) scores 48.5% on ExploitGym in OpenAI's chart, against 32.1% for this model; no independent benchmark score is recorded here"
      ],
      "bestFor": [
        "Daybreak Red teams that need exploit validation and security testing at a lower price than GPT-6 Cyber"
      ],
      "pricing": {
        "inputPerMillionUsd": 12.5,
        "outputPerMillionUsd": 75,
        "cachedInputPerMillionUsd": 1.25,
        "sourceId": "openai-gpt-5-6-cyber-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 12.5,
          "tier": "standard",
          "sourceId": "openai-gpt-5-6-cyber-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": 75,
          "tier": "standard",
          "sourceId": "openai-gpt-5-6-cyber-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 1.25,
          "tier": "cached",
          "sourceId": "openai-gpt-5-6-cyber-model"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "openai-daybreak-expansion",
        "openai-gpt-5-6-cyber-model"
      ]
    },
    {
      "id": "gpt-image-2-6-sunburst",
      "providerId": "openai",
      "name": "GPT Image 2.6 Sunburst",
      "family": "GPT Image",
      "releaseDate": "2026-11-04",
      "status": "active",
      "access": "api",
      "apiModelId": "gpt-image-2.6-sunburst",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "image"
      ],
      "capabilities": [
        "image-generation",
        "image-editing",
        "image-to-image"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "image"
      ],
      "strengths": [
        "OpenAI's higher-quality model in Images 2.6, with more consistent characters, products and brand details across edits at medium or high quality",
        "Makes animated GIFs and SVG files, takes a complexity setting for how much the model plans a scene and supports markup-guided edits; OpenAI says Images 2.6 generates 50% faster than before",
        "Token rates match GPT Image 2.5: $5 per 1M text input tokens, $8 per 1M image input tokens and $30 per 1M image output tokens"
      ],
      "cautions": [
        "Arena's first ratings put it below GPT Image 2.5 Sunburst: 1409 against 1427 for generation and 1476 against 1517 for editing, on fewer than 4,000 votes so far",
        "OpenAI's pricing page lists no batch rate for 2.6 yet, while GPT Image 2.5 costs half in batch"
      ],
      "bestFor": [
        "Detailed compositions, brand-consistent edits and animated GIFs where quality matters most"
      ],
      "pricing": {
        "inputPerMillionUsd": 5,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": 1.25,
        "sourceId": "openai-gpt-image-2-6-sunburst-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 5,
          "tier": "standard",
          "sourceId": "openai-gpt-image-2-6-sunburst-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "openai-gpt-image-2-6-sunburst-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 1.25,
          "tier": "cached",
          "sourceId": "openai-gpt-image-2-6-sunburst-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 8,
          "tier": "standard-image-input",
          "sourceId": "openai-gpt-image-2-6-sunburst-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": 30,
          "tier": "standard-image-output",
          "sourceId": "openai-gpt-image-2-6-sunburst-model"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-image",
          "benchmarkSnapshotId": "arena-text-to-image-snapshot-2026-11-04",
          "score": 1409,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "GPT-Image-2.6 Sunburst (Medium)"
        },
        {
          "benchmarkId": "arena-image-edit",
          "benchmarkSnapshotId": "arena-image-edit-snapshot-2026-11-04",
          "score": 1476,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "GPT-Image-2.6 Sunburst (Medium)"
        }
      ],
      "officialSourceIds": [
        "openai-gpt-image-2-6-sunburst-model",
        "openai-images-2-6"
      ]
    },
    {
      "id": "gpt-image-2-6-flare",
      "providerId": "openai",
      "name": "GPT Image 2.6 Flare",
      "family": "GPT Image",
      "releaseDate": "2026-11-04",
      "status": "active",
      "access": "api",
      "apiModelId": "gpt-image-2.6-flare",
      "contextWindowTokens": null,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "image"
      ],
      "capabilities": [
        "image-generation",
        "image-editing",
        "image-to-image"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "image"
      ],
      "strengths": [
        "OpenAI's faster model in Images 2.6: as little as six seconds per image at low quality, with better layouts for posters and product imagery",
        "Makes animated GIFs and SVG files, takes a complexity setting for how much the model plans a scene and supports markup-guided edits; OpenAI says Images 2.6 generates 50% faster than before",
        "Token rates match GPT Image 2.5: $5 per 1M text input tokens, $8 per 1M image input tokens and $30 per 1M image output tokens"
      ],
      "cautions": [
        "Arena's first ratings put it below GPT Image 2.5 Flare: 1384 against 1399 for generation and 1462 against 1482 for editing, on fewer than 4,000 votes so far",
        "OpenAI's pricing page lists no batch rate for 2.6 yet, while GPT Image 2.5 costs half in batch"
      ],
      "bestFor": [
        "Fast, high-volume image generation and quick edits"
      ],
      "pricing": {
        "inputPerMillionUsd": 5,
        "outputPerMillionUsd": null,
        "cachedInputPerMillionUsd": 1.25,
        "sourceId": "openai-gpt-image-2-6-flare-model"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 5,
          "tier": "standard",
          "sourceId": "openai-gpt-image-2-6-flare-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": null,
          "tier": "standard",
          "sourceId": "openai-gpt-image-2-6-flare-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 1.25,
          "tier": "cached",
          "sourceId": "openai-gpt-image-2-6-flare-model"
        },
        {
          "unit": "million-input-tokens",
          "usd": 8,
          "tier": "standard-image-input",
          "sourceId": "openai-gpt-image-2-6-flare-model"
        },
        {
          "unit": "million-output-tokens",
          "usd": 30,
          "tier": "standard-image-output",
          "sourceId": "openai-gpt-image-2-6-flare-model"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-image",
          "benchmarkSnapshotId": "arena-text-to-image-snapshot-2026-11-04",
          "score": 1384,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "GPT-Image-2.6 Flare (Medium)"
        },
        {
          "benchmarkId": "arena-image-edit",
          "benchmarkSnapshotId": "arena-image-edit-snapshot-2026-11-04",
          "score": 1462,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-04",
          "sourceRow": "GPT-Image-2.6 Flare (Medium)"
        }
      ],
      "officialSourceIds": [
        "openai-gpt-image-2-6-flare-model",
        "openai-images-2-6"
      ]
    },
    {
      "id": "gemini-omni-1-2-flash",
      "providerId": "google",
      "name": "Gemini Omni 1.2 Flash",
      "family": "Gemini Omni",
      "releaseDate": "2026-11-05",
      "status": "active",
      "access": "api",
      "apiModelId": "gemini-omni-1.2-flash",
      "contextWindowTokens": 131072,
      "maxOutputTokens": 65536,
      "modalities": [
        "text",
        "image",
        "video",
        "audio"
      ],
      "taskFamilies": [
        "video"
      ],
      "capabilities": [
        "text-to-video",
        "image-to-video",
        "video-editing"
      ],
      "inputModalities": [
        "text",
        "image",
        "video"
      ],
      "outputModalities": [
        "video",
        "audio",
        "text"
      ],
      "strengths": [
        "Google's newest conversational video model, generally available on the paid tier: clips of 3 to 15 seconds with native audio, refined over several turns through the Interactions API",
        "Arena rates it first on both its text-to-video (1541) and image-to-video (1502) leaderboards (November 5, 2026), above Gemini Omni 1.1 Flash",
        "Video output costs $12 per 1M tokens, 30% less than Gemini Omni 1.1 Flash: about $0.07 a second at 720p, $0.10 at 1080p and $0.21 at 4K",
        "Keeps characters and subjects consistent across up to 6 reference images, and extends a clip by 3 to 10 seconds a turn up to 30 seconds"
      ],
      "cautions": [
        "Arena's ratings rest on 2,745 votes or fewer per board so far, far fewer than older video models",
        "1080p and 4K come from upsampling, and first-and-last-frame interpolation is limited to 3 to 10 seconds",
        "Paid tier only: there is no free-tier access in the Gemini API"
      ],
      "bestFor": [
        "Short clips with sound and consistent characters that need conversational edits"
      ],
      "pricing": {
        "inputPerMillionUsd": 1.5,
        "outputPerMillionUsd": 9,
        "cachedInputPerMillionUsd": null,
        "sourceId": "google-pricing-omni-1-2-flash"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 1.5,
          "tier": "standard",
          "sourceId": "google-pricing-omni-1-2-flash"
        },
        {
          "unit": "million-output-tokens",
          "usd": 9,
          "tier": "standard",
          "sourceId": "google-pricing-omni-1-2-flash"
        },
        {
          "unit": "million-input-tokens",
          "usd": null,
          "tier": "cached",
          "sourceId": "google-pricing-omni-1-2-flash"
        },
        {
          "unit": "million-output-tokens",
          "usd": 12,
          "tier": "video-output",
          "sourceId": "google-pricing-omni-1-2-flash"
        },
        {
          "unit": "video-second",
          "usd": 0.07,
          "tier": "standard-720p",
          "sourceId": "google-pricing-omni-1-2-flash"
        }
      ],
      "benchmarkResults": [
        {
          "benchmarkId": "arena-text-to-video",
          "benchmarkSnapshotId": "arena-text-to-video-snapshot-2026-11-05",
          "score": 1541,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-05",
          "sourceRow": "Gemini Omni 1.2 Flash"
        },
        {
          "benchmarkId": "arena-image-to-video",
          "benchmarkSnapshotId": "arena-image-to-video-snapshot-2026-11-05",
          "score": 1502,
          "sourceId": "arena-leaderboard-dataset-2026-11-06",
          "measuredAt": "2026-11-05",
          "sourceRow": "Gemini Omni 1.2 Flash"
        }
      ],
      "officialSourceIds": [
        "google-gemini-omni-flash-model",
        "google-gemini-changelog-2026-11-05",
        "google-pricing-omni-1-2-flash"
      ]
    },
    {
      "id": "inkling-2-small",
      "providerId": "thinking-machines",
      "name": "Inkling 2 Small",
      "family": "Inkling",
      "releaseDate": "2026-11-05",
      "status": "active",
      "access": "api",
      "apiModelId": "thinkingmachines/Inkling-2-Small",
      "contextWindowTokens": 262144,
      "maxOutputTokens": null,
      "modalities": [
        "text",
        "image"
      ],
      "taskFamilies": [
        "text-nlp",
        "multimodal"
      ],
      "capabilities": [
        "reasoning",
        "coding",
        "agentic-coding",
        "image-understanding"
      ],
      "inputModalities": [
        "text",
        "image"
      ],
      "outputModalities": [
        "text"
      ],
      "strengths": [
        "Thinking Machines Lab's second model generation: a mixture-of-experts model with 594B total and 17B active parameters that reads images and takes a variable effort setting",
        "In its lab's table no other listed model scores higher on EnterpriseBench-CoreCraft (71.8%), Finance Agent v2.1 (67.5%) or ScreenSpot-Pro (83.8%)",
        "$0.30 input, $1.00 output and $0.03 cached input per 1M tokens on Tinker, which can also fine-tune it with RL or supervised training"
      ],
      "cautions": [
        "Closed weights for now: Thinking Machines says it will publish the full weights under Apache 2.0 the week after launch, once external safety testing ends; until then Tinker is the only way to use it",
        "Tinker serves it with a 64K or 256K context, although the model card lists up to 1M tokens",
        "In its lab's own table DeepSeek-V4.1-Flash scores higher on DeepSWE v1.1 (77.0% against 72.1%) and Claude Haiku 5.5 on Vibe Code Bench (90.4% against 81.6%); Thinking Machines also says it is less tuned for chat, and no independent benchmark score is recorded here yet"
      ],
      "bestFor": [
        "Agentic coding and office, finance and legal work at a low per-token price",
        "Teams that want to post-train their own variant on Tinker"
      ],
      "pricing": {
        "inputPerMillionUsd": 0.3,
        "outputPerMillionUsd": 1,
        "cachedInputPerMillionUsd": 0.03,
        "sourceId": "tinker-models-pricing"
      },
      "prices": [
        {
          "unit": "million-input-tokens",
          "usd": 0.3,
          "tier": "standard",
          "sourceId": "tinker-models-pricing"
        },
        {
          "unit": "million-output-tokens",
          "usd": 1,
          "tier": "standard",
          "sourceId": "tinker-models-pricing"
        },
        {
          "unit": "million-input-tokens",
          "usd": 0.03,
          "tier": "cached",
          "sourceId": "tinker-models-pricing"
        }
      ],
      "benchmarkResults": [],
      "officialSourceIds": [
        "thinking-machines-inkling-2-small-release",
        "thinking-machines-inkling-2-small-card",
        "tinker-models-pricing"
      ]
    }
  ],
  "sources": [
    {
      "id": "livebench-2026-06-25-averaging-19e766a5",
      "title": "LiveBench category-weighted averaging implementation",
      "publisher": "LiveBench",
      "url": "https://github.com/LiveBench/livebench.github.io/blob/19e766a5de4de07d672ed5bf9f0a69ceed1d39bf/src/Table/Averaging.js",
      "sourceType": "independent",
      "verifiedAt": "2026-08-07",
      "datasetVersion": "LiveBench-2026-06-25@19e766a5de4de07d672ed5bf9f0a69ceed1d39bf",
      "license": "Apache-2.0",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "Pinned implementation averages task scores within each of the seven categories, then averages the category scores; the paired categories_2026_06_25.json file defines category membership."
    },
    {
      "id": "livebench-2026-06-25-results-19e766a5",
      "title": "LiveBench 2026-06-25 task results",
      "publisher": "LiveBench",
      "url": "https://github.com/LiveBench/livebench.github.io/blob/19e766a5de4de07d672ed5bf9f0a69ceed1d39bf/public/table_2026_06_25.csv",
      "sourceType": "independent",
      "verifiedAt": "2026-08-07",
      "datasetVersion": "LiveBench-2026-06-25@19e766a5de4de07d672ed5bf9f0a69ceed1d39bf",
      "license": "Apache-2.0",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "Snapshot and measurement date: 2026-06-25. Overall scores are imported only for unambiguous model mappings and are recomputed with the pinned category-weighted averaging implementation."
    },
    {
      "id": "openai-gpt-5-4-model",
      "title": "GPT-5.4 model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-5.4",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official model documentation verifies the dated snapshot, token limits, modalities, and exact standard token prices represented here."
    },
    {
      "id": "openai-gpt-5-5-release",
      "title": "Introducing GPT-5.5",
      "publisher": "OpenAI",
      "url": "https://openai.com/index/introducing-gpt-5-5/",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official release announcement verifies model identity and the 2026-04-23 release date."
    },
    {
      "id": "openai-gpt-5-5-model",
      "title": "GPT-5.5 model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-5.5",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official model documentation verifies token limits, modalities, and exact standard token prices represented here."
    },
    {
      "id": "openai-gpt-5-6-release",
      "title": "GPT-5.6: Frontier intelligence that scales with your ambition",
      "publisher": "OpenAI",
      "url": "https://openai.com/index/gpt-5-6/",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official release announcement verifies the Sol, Terra, and Luna identities and 2026-07-09 general-availability date."
    },
    {
      "id": "openai-gpt-5-6-models",
      "title": "OpenAI GPT-5.6 model catalog",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official model catalog verifies the Sol, Terra, and Luna token limits, modalities, roles, and exact standard token prices represented here."
    },
    {
      "id": "anthropic-claude-opus-4-8-release",
      "title": "Introducing Claude Opus 4.8",
      "publisher": "Anthropic",
      "url": "https://www.anthropic.com/news/claude-opus-4-8",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official release documentation verifies model identity, release date, the published context window, and standard input/output pricing."
    },
    {
      "id": "anthropic-claude-sonnet-5-release",
      "title": "Introducing Claude Sonnet 5",
      "publisher": "Anthropic",
      "url": "https://www.anthropic.com/news/claude-sonnet-5",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official release documentation verifies model identity, release date, API availability, and introductory pricing through 2026-08-31."
    },
    {
      "id": "anthropic-claude-fable-5-release",
      "title": "Claude Fable 5 and Claude Mythos 5",
      "publisher": "Anthropic",
      "url": "https://www.anthropic.com/news/claude-fable-5-mythos-5",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official release announcement verifies Claude Fable 5 identity, 2026-06-09 release date, API access, standard token pricing, and general availability after redeployment."
    },
    {
      "id": "anthropic-models-overview-current",
      "title": "Claude models overview",
      "publisher": "Anthropic",
      "url": "https://platform.claude.com/docs/en/about-claude/models/overview",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official current-model documentation verifies Fable's text-and-image inputs, 1,000,000-token context window, 128,000-token output limit, and current API availability."
    },
    {
      "id": "google-gemini-3-5-changelog",
      "title": "Gemini API release notes",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/changelog",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Official release notes verify the dated Gemini 3.5 Flash, Gemini 3.1 Flash Live, Gemini 3.1 Flash TTS, Gemini 3.1 Flash Lite Image, Gemini 3 Pro Image, and Gemini Omni Flash releases represented in this catalog."
    },
    {
      "id": "google-gemini-changelog-2026-10-08",
      "title": "Gemini API release notes: October 8, 2026",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/changelog#10-08-2026",
      "sourceType": "official",
      "verifiedAt": "2026-10-29",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Release note verifies that gemini-3.7-flash and gemini-3.5-flash were deprecated on 2026-10-08 and replaced by gemini-3.8-flash and gemini-3.6-flash, that Google AI Studio no longer offers them and that API requests to them are routed to their replacements."
    },
    {
      "id": "google-gemini-changelog-2026-10-06",
      "title": "Gemini API release notes: October 6, 2026",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/changelog#10-06-2026",
      "sourceType": "official",
      "verifiedAt": "2026-10-29",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Release note verifies that gemini-3.1-flash-image was deprecated on 2026-10-06 with no shutdown date announced, and that Google recommends migrating to gemini-nano-banana-2.1."
    },
    {
      "id": "google-gemini-3-5-model",
      "title": "Gemini 3.5 Flash model",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models/gemini-3.5-flash",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Official model documentation verifies model code, modalities, and token limits; exact token prices remain unreported here."
    },
    {
      "id": "google-gemini-3-6-model",
      "title": "Gemini 3.6 Flash model",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models/gemini-3.6-flash",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Official model documentation verifies model code, 2026-07-21 update, modalities, and token limits."
    },
    {
      "id": "google-gemini-3-6-latest",
      "title": "Using the latest Gemini models",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/latest-model",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official latest-model guidance verifies the exact standard input/output prices represented for Gemini 3.6 Flash."
    },
    {
      "id": "google-gemini-3-5-flash-lite-model",
      "title": "Gemini 3.5 Flash-Lite model",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models/gemini-3.5-flash-lite",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Official model documentation verifies the stable model code, multimodal inputs, token limits, capabilities, and July 2026 update."
    },
    {
      "id": "google-gemini-3-5-flash-lite-pricing",
      "title": "Gemini 3.5 Flash-Lite pricing",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3.5-flash-lite",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "pricing"
      ],
      "reuseNote": "Official pricing table verifies the standard paid-tier input, cached-input, and output rates represented here."
    },
    {
      "id": "xai-grok-4-3-model",
      "title": "Grok 4.3 model",
      "publisher": "SpaceXAI",
      "url": "https://docs.x.ai/developers/models/grok-4.3",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official model documentation verifies Grok 4.3 modalities, context window, reasoning modes, aliases, and standard short-context prices."
    },
    {
      "id": "xai-inference-models-api",
      "title": "xAI inference models API reference",
      "publisher": "SpaceXAI",
      "url": "https://docs.x.ai/developers/rest-api-reference/inference/models",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official API reference exposes Grok 4.3 modalities, aliases, and token pricing fields; its object creation timestamp is not treated as a public release date."
    },
    {
      "id": "xai-grok-4-3-public-availability",
      "title": "Skills in web, iOS, and Android",
      "publisher": "SpaceXAI",
      "url": "https://x.ai/news/grok-skills",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official 2026-05-18 announcement confirms Grok 4.3 was live across grok.com, iOS, and Android by that date; it is not treated as the model's exact release date."
    },
    {
      "id": "xai-release-notes-current",
      "title": "SpaceXAI API release notes",
      "publisher": "SpaceXAI",
      "url": "https://docs.x.ai/developers/release-notes",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official release notes verify Grok 4.20 and Grok 4.20 Multi-agent API availability on 2026-03-10."
    },
    {
      "id": "xai-grok-4-20-system-card",
      "title": "Grok 4.20 System Card",
      "publisher": "SpaceXAI",
      "url": "https://data.x.ai/2026-04-07-grok-4-20-model-card.pdf",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": "2026-04-07",
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official system card verifies Grok 4.20 identity, single-agent and multi-agent deployment modes, supported input modalities, and intended uses."
    },
    {
      "id": "xai-current-model-pricing",
      "title": "SpaceXAI model pricing",
      "publisher": "SpaceXAI",
      "url": "https://docs.x.ai/developers/pricing",
      "sourceType": "official",
      "verifiedAt": "2026-08-14",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official current pricing table verifies active Grok 4.6, Grok 4.5, and Grok 4.20 variants, their documented context windows, and standard and long-context token rates."
    },
    {
      "id": "xai-may-15-retirement",
      "title": "Grok model retirement on May 15, 2026",
      "publisher": "SpaceXAI",
      "url": "https://docs.x.ai/developers/migration/may-15-retirement",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official lifecycle notice verifies Grok 3 retirement and redirect to Grok 4.3, and identifies Grok 4.3 as the recommended general replacement."
    },
    {
      "id": "xai-grok-4-5-release",
      "title": "Introducing Grok 4.5",
      "publisher": "SpaceXAI",
      "url": "https://x.ai/news/grok-4-5",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official announcement verifies the model identity, release date, and intended coding, agentic, and knowledge-work scope."
    },
    {
      "id": "xai-grok-4-5-model",
      "title": "Grok 4.5 model",
      "publisher": "SpaceXAI",
      "url": "https://docs.x.ai/developers/models/grok-4.5",
      "sourceType": "official",
      "verifiedAt": "2026-08-07",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official model documentation verifies modalities, context window, and standard short-context token prices; the model caution discloses higher long-context rates."
    },
    {
      "id": "xai-grok-4-6-release",
      "title": "Introducing Grok 4.6",
      "publisher": "SpaceXAI",
      "url": "https://x.ai/news/grok-4-6",
      "sourceType": "official",
      "verifiedAt": "2026-08-14",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official announcement verifies the 2026-08-12 release date and positioning for long-running agents, coding, and knowledge work."
    },
    {
      "id": "xai-grok-4-6-model",
      "title": "Grok 4.6 model",
      "publisher": "SpaceXAI",
      "url": "https://docs.x.ai/developers/models/grok-4.6",
      "sourceType": "official",
      "verifiedAt": "2026-08-14",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official model documentation verifies the exact model ID, text and image input, text output, 500,000-token context window, reasoning, function calling, structured outputs, and standard token prices."
    },
    {
      "id": "livebench-2024-11-25-methodology",
      "title": "LiveBench datasheet and methodology",
      "publisher": "LiveBench",
      "url": "https://github.com/LiveBench/LiveBench/blob/00eae856aa1c1a9e9d058a65a9a94d85884034c4/docs/DATASHEET.md",
      "sourceType": "independent",
      "verifiedAt": "2026-08-06",
      "datasetVersion": "LiveBench-2024-11-25",
      "license": "Apache-2.0",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "Primary benchmark documentation confirms objective scoring, public distribution, and Apache 2.0 reuse terms."
    },
    {
      "id": "livebench-2024-11-25-results-347e3d0b",
      "title": "LiveBench 2024-11-25 task results",
      "publisher": "LiveBench",
      "url": "https://github.com/LiveBench/livebench.github.io/blob/347e3d0b6a4916cd66f8bda31ce7ecdd7436ecc3/public/table_2024_11_25.csv",
      "sourceType": "independent",
      "verifiedAt": "2026-08-06",
      "datasetVersion": "LiveBench-2024-11-25@347e3d0b6a4916cd66f8bda31ce7ecdd7436ecc3",
      "license": "Apache-2.0",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "Snapshot and measurement date: 2024-11-25. Exact published snapshot; only direct Web of Lies V2 percentages with unambiguous model mappings are imported."
    },
    {
      "id": "openai-gpt-4o-release",
      "title": "Hello GPT-4o",
      "publisher": "OpenAI",
      "url": "https://openai.com/index/hello-gpt-4o/",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "openai-gpt-4-1-release",
      "title": "Introducing GPT-4.1 in the API",
      "publisher": "OpenAI",
      "url": "https://openai.com/index/gpt-4-1/",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "openai-pricing",
      "title": "OpenAI API pricing",
      "publisher": "OpenAI",
      "url": "https://openai.com/api/pricing/",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "pricing"
      ],
      "reuseNote": "Official pricing page used only for the exact token prices represented in model records."
    },
    {
      "id": "anthropic-claude-3-release",
      "title": "Introducing the next generation of Claude",
      "publisher": "Anthropic",
      "url": "https://www.anthropic.com/news/claude-3-family",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "anthropic-claude-3-7-release",
      "title": "Claude 3.7 Sonnet and Claude Code",
      "publisher": "Anthropic",
      "url": "https://www.anthropic.com/news/claude-3-7-sonnet",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "anthropic-pricing",
      "title": "Claude pricing",
      "publisher": "Anthropic",
      "url": "https://docs.anthropic.com/en/docs/about-claude/pricing",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "pricing"
      ],
      "reuseNote": "Official pricing documentation used only for the exact token prices represented in model records."
    },
    {
      "id": "google-gemini-1-5-release",
      "title": "Our next-generation model: Gemini 1.5",
      "publisher": "Google",
      "url": "https://blog.google/technology/ai/google-gemini-next-generation-model-february-2024/",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "google-gemini-1-5-flash-release",
      "title": "Gemini 1.5 Flash",
      "publisher": "Google",
      "url": "https://developers.googleblog.com/en/gemini-15-flash-updates-google-ai-studio-gemini-api/",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "google-gemini-2-release",
      "title": "Introducing Gemini 2.0",
      "publisher": "Google",
      "url": "https://blog.google/technology/google-deepmind/google-gemini-ai-update-december-2024/",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "google-gemini-2-5-release",
      "title": "Gemini 2.5: Our most intelligent AI model",
      "publisher": "Google",
      "url": "https://blog.google/technology/google-deepmind/gemini-model-thinking-updates-march-2025/",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "google-pricing",
      "title": "Gemini Developer API pricing",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/pricing",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "pricing"
      ],
      "reuseNote": "Official Gemini Developer API pricing table used only for exact published token, image, audio-minute, and video-second prices represented in model records."
    },
    {
      "id": "meta-llama-3-release",
      "title": "Introducing Meta Llama 3",
      "publisher": "Meta",
      "url": "https://ai.meta.com/blog/meta-llama-3/",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "meta-llama-3-1-release",
      "title": "Introducing Llama 3.1",
      "publisher": "Meta",
      "url": "https://ai.meta.com/blog/meta-llama-3-1/",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "meta-llama-3-2-release",
      "title": "Llama 3.2",
      "publisher": "Meta",
      "url": "https://ai.meta.com/blog/llama-3-2-connect-2024-vision-edge-mobile-devices/",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "xai-grok-1-5-release",
      "title": "Announcing Grok-1.5",
      "publisher": "xAI",
      "url": "https://x.ai/news/grok-1.5",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "xai-grok-1-5v-release",
      "title": "Grok-1.5 Vision Preview",
      "publisher": "xAI",
      "url": "https://x.ai/news/grok-1.5v",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "xai-grok-2-release",
      "title": "Grok-2 Beta Release",
      "publisher": "xAI",
      "url": "https://x.ai/news/grok-2",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "xai-grok-3-release",
      "title": "Grok 3 Beta",
      "publisher": "xAI",
      "url": "https://x.ai/news/grok-3",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "xai-models-pricing",
      "title": "xAI models and pricing",
      "publisher": "xAI",
      "url": "https://docs.x.ai/docs/models",
      "sourceType": "official",
      "verifiedAt": "2026-10-11",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official model catalog used to verify documented model availability and pricing fields, not historical release dates."
    },
    {
      "id": "deepseek-v2-release",
      "title": "DeepSeek-V2",
      "publisher": "DeepSeek",
      "url": "https://github.com/deepseek-ai/DeepSeek-V2",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "deepseek-coder-v2-release",
      "title": "DeepSeek-Coder-V2",
      "publisher": "DeepSeek",
      "url": "https://github.com/deepseek-ai/DeepSeek-Coder-V2",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "deepseek-v3-release",
      "title": "DeepSeek-V3",
      "publisher": "DeepSeek",
      "url": "https://github.com/deepseek-ai/DeepSeek-V3",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "deepseek-r1-release",
      "title": "DeepSeek-R1",
      "publisher": "DeepSeek",
      "url": "https://github.com/deepseek-ai/DeepSeek-R1",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "deepseek-pricing",
      "title": "DeepSeek API pricing",
      "publisher": "DeepSeek",
      "url": "https://api-docs.deepseek.com/quick_start/pricing",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "pricing"
      ],
      "reuseNote": "Official pricing documentation consulted for price provenance; unavailable exact values remain unreported."
    },
    {
      "id": "mistral-codestral-release",
      "title": "Codestral",
      "publisher": "Mistral AI",
      "url": "https://mistral.ai/news/codestral/",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "mistral-large-2-release",
      "title": "Large Enough",
      "publisher": "Mistral AI",
      "url": "https://mistral.ai/news/mistral-large-2407/",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "mistral-pixtral-release",
      "title": "Announcing Pixtral 12B",
      "publisher": "Mistral AI",
      "url": "https://mistral.ai/news/pixtral-12b/",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "mistral-ministral-release",
      "title": "Un Ministral, des Ministraux",
      "publisher": "Mistral AI",
      "url": "https://mistral.ai/news/ministraux/",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "mistral-pricing",
      "title": "Mistral AI pricing",
      "publisher": "Mistral AI",
      "url": "https://mistral.ai/products/la-plateforme#pricing",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "pricing"
      ],
      "reuseNote": "Official pricing page consulted for price provenance; unavailable exact values remain unreported."
    },
    {
      "id": "qwen2-release",
      "title": "Hello Qwen2",
      "publisher": "Qwen Team",
      "url": "https://qwenlm.github.io/blog/qwen2/",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "qwen2-5-release",
      "title": "Qwen2.5: A Party of Foundation Models",
      "publisher": "Qwen Team",
      "url": "https://qwenlm.github.io/blog/qwen2.5/",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "qwen2-5-vl-release",
      "title": "Qwen2.5-VL",
      "publisher": "Qwen Team",
      "url": "https://qwenlm.github.io/blog/qwen2.5-vl/",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "kimi-vl-release",
      "title": "Kimi-VL",
      "publisher": "Moonshot AI",
      "url": "https://github.com/MoonshotAI/Kimi-VL",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "kimi-dev-release",
      "title": "Kimi-Dev",
      "publisher": "Moonshot AI",
      "url": "https://github.com/MoonshotAI/Kimi-Dev",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "kimi-k2-release",
      "title": "Kimi K2",
      "publisher": "Moonshot AI",
      "url": "https://github.com/MoonshotAI/Kimi-K2",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify model identity and release details."
    },
    {
      "id": "openai-gpt-4o-model",
      "title": "GPT-4o model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-4o",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official provider documentation used to verify exact model specifications."
    },
    {
      "id": "openai-gpt-4-1-model",
      "title": "GPT-4.1 model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-4.1",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official provider documentation used to verify exact model specifications."
    },
    {
      "id": "openai-gpt-4-1-mini-model",
      "title": "GPT-4.1 mini model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-4.1-mini",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official provider documentation used to verify exact model specifications."
    },
    {
      "id": "openai-gpt-4-1-nano-model",
      "title": "GPT-4.1 nano model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-4.1-nano",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official provider documentation used to verify exact model specifications."
    },
    {
      "id": "anthropic-claude-3-haiku-release",
      "title": "Claude 3 Haiku",
      "publisher": "Anthropic",
      "url": "https://www.anthropic.com/news/claude-3-haiku",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify exact model specifications."
    },
    {
      "id": "anthropic-models-overview",
      "title": "Claude models overview",
      "publisher": "Anthropic",
      "url": "https://docs.anthropic.com/en/docs/about-claude/models/overview",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify exact model specifications."
    },
    {
      "id": "google-models-overview",
      "title": "Gemini API models",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify exact model specifications."
    },
    {
      "id": "google-gemini-2-5-pro-model",
      "title": "Gemini 2.5 Pro",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models/gemini-2.5-pro",
      "sourceType": "official",
      "verifiedAt": "2026-08-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official provider documentation used to verify exact model specifications."
    },
    {
      "id": "specialist-openai-gpt-image-2",
      "title": "GPT Image 2 model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-image-2",
      "sourceType": "official",
      "verifiedAt": "2026-08-10",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "specialist-openai-realtime-1-5",
      "title": "GPT-Realtime-1.5 model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-realtime-1.5",
      "sourceType": "official",
      "verifiedAt": "2026-08-10",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "specialist-openai-audio-1-5",
      "title": "gpt-audio-1.5 model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-audio-1.5",
      "sourceType": "official",
      "verifiedAt": "2026-08-10",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "specialist-openai-transcribe",
      "title": "GPT-4o Transcribe model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-4o-transcribe",
      "sourceType": "official",
      "verifiedAt": "2026-08-10",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "specialist-openai-tts-1",
      "title": "TTS-1 model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/tts-1",
      "sourceType": "official",
      "verifiedAt": "2026-08-10",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "specialist-google-gemini-3-1-flash-image-pricing",
      "title": "Gemini 3.1 Flash Image pricing",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3.1-flash-image-nano-banana-2",
      "sourceType": "official",
      "verifiedAt": "2026-08-10",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party pricing table used to verify Gemini 3.1 Flash Image standard token and resolution-specific image prices."
    },
    {
      "id": "specialist-google-native-image-capabilities",
      "title": "Gemini native image generation capabilities",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/image-generation",
      "sourceType": "official",
      "verifiedAt": "2026-08-10",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party native image guide used to verify Gemini 3.1 Flash Image generation, editing, image input, and mixed text/image output capabilities."
    },
    {
      "id": "specialist-google-native-media",
      "title": "Gemini Developer API pricing and native media model identifiers",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/pricing#veo-3.1",
      "sourceType": "official",
      "verifiedAt": "2026-08-10",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "specialist-xai-imagine-image",
      "title": "grok-imagine-image model",
      "publisher": "SpaceXAI",
      "url": "https://docs.x.ai/developers/models/grok-imagine-image",
      "sourceType": "official",
      "verifiedAt": "2026-08-10",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "specialist-xai-imagine-video",
      "title": "Video generation",
      "publisher": "SpaceXAI",
      "url": "https://docs.x.ai/developers/model-capabilities/video/generation",
      "sourceType": "official",
      "verifiedAt": "2026-08-10",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "specialist-xai-imagine-pricing",
      "title": "Imagine API pricing",
      "publisher": "SpaceXAI",
      "url": "https://docs.x.ai/developers/pricing#imagine-api",
      "sourceType": "official",
      "verifiedAt": "2026-08-10",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "specialist-runway-models",
      "title": "Available AI Models",
      "publisher": "Runway",
      "url": "https://docs.dev.runwayml.com/guides/models/",
      "sourceType": "official",
      "verifiedAt": "2026-08-10",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "specialist-elevenlabs-models",
      "title": "ElevenLabs models",
      "publisher": "ElevenLabs",
      "url": "https://elevenlabs.io/docs/overview/models",
      "sourceType": "official",
      "verifiedAt": "2026-08-10",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "specialist-elevenlabs-scribe-v2-release",
      "title": "Introducing Scribe v2",
      "publisher": "ElevenLabs",
      "url": "https://elevenlabs.io/blog/introducing-scribe-v2",
      "sourceType": "official",
      "verifiedAt": "2026-08-10",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "specialist-elevenlabs-scribe-realtime-release",
      "title": "Introducing Scribe v2 Realtime",
      "publisher": "ElevenLabs",
      "url": "https://elevenlabs.io/blog/introducing-scribe-v2-realtime",
      "sourceType": "official",
      "verifiedAt": "2026-08-10",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "specialist-stability-pricing",
      "title": "Stability AI Developer Platform pricing",
      "publisher": "Stability AI",
      "url": "https://platform.stability.ai/pricing",
      "sourceType": "official",
      "verifiedAt": "2026-08-10",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "specialist-bfl-models",
      "title": "FLUX image generation endpoints",
      "publisher": "Black Forest Labs",
      "url": "https://docs.bfl.ai/quick_start/generating_images",
      "sourceType": "official",
      "verifiedAt": "2026-08-10",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "specialist-bfl-pricing",
      "title": "FLUX API pricing",
      "publisher": "Black Forest Labs",
      "url": "https://docs.bfl.ai/quick_start/pricing",
      "sourceType": "official",
      "verifiedAt": "2026-08-10",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "anthropic-pricing-2026-09",
      "title": "Claude API pricing",
      "publisher": "Anthropic",
      "url": "https://platform.claude.com/docs/en/about-claude/pricing",
      "sourceType": "official",
      "verifiedAt": "2026-10-11",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "anthropic-models-overview-2026-09",
      "title": "Claude models overview",
      "publisher": "Anthropic",
      "url": "https://platform.claude.com/docs/en/models/overview",
      "sourceType": "official",
      "verifiedAt": "2026-10-11",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "anthropic-glasswing-2026-09",
      "title": "Claude Mythos limited availability",
      "publisher": "Anthropic",
      "url": "https://www.anthropic.com/glasswing",
      "sourceType": "official",
      "verifiedAt": "2026-09-09",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "openai-gpt-6-astra-model",
      "title": "GPT-6 Astra model reference",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-6-astra",
      "sourceType": "official",
      "verifiedAt": "2026-09-09",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "openai-api-pricing-2026-09",
      "title": "OpenAI API pricing",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/pricing",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "google-gemini-38-flash-2026-09",
      "title": "Gemini 3.8 Flash model card",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models/gemini-3.8-flash",
      "sourceType": "official",
      "verifiedAt": "2026-09-09",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "meta-model-api-models-2026-09",
      "title": "Meta Model API model list",
      "publisher": "Meta",
      "url": "https://dev.meta.ai/docs/models",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "kimi-platform-pricing-2026-09",
      "title": "Kimi platform model and pricing documentation",
      "publisher": "Moonshot AI",
      "url": "https://platform.kimi.ai/docs/pricing/chat",
      "sourceType": "official",
      "verifiedAt": "2026-10-28",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "alibaba-model-studio-models-2026-09",
      "title": "Alibaba Cloud Model Studio model list",
      "publisher": "Alibaba Cloud",
      "url": "https://www.alibabacloud.com/help/en/model-studio/models",
      "sourceType": "official",
      "verifiedAt": "2026-09-09",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "openrouter-models-api-2026-09",
      "title": "OpenRouter models API catalogue",
      "publisher": "OpenRouter",
      "url": "https://openrouter.ai/api/v1/models",
      "sourceType": "independent",
      "verifiedAt": "2026-09-09",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Third-party API catalogue used only to cross-check published token prices, context limits, and listing dates against first-party documentation."
    },
    {
      "id": "google-gemini-3-7-flash-model",
      "title": "Gemini 3.7 Flash model",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models/gemini-3.7-flash",
      "sourceType": "official",
      "verifiedAt": "2026-10-29",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Official model documentation verifies the Gemini 3.7 Flash code and its deprecation notice (API requests route to gemini-3.8-flash), text/image/video/audio inputs, text output, token limits, structured outputs, thinking, and URL context."
    },
    {
      "id": "google-gemini-3-7-flash-release",
      "title": "Introducing Gemini 3.7 Flash",
      "publisher": "Google",
      "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-3-7-flash/",
      "sourceType": "official",
      "verifiedAt": "2026-08-18",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Official Google announcement verifies the 2026-08-13 release, coding and agent positioning, and introductory token prices through 2026-12-31 before the documented 2027 rate change."
    },
    {
      "id": "google-gemini-3-1-pro-preview-model",
      "title": "Gemini 3.1 Pro Preview",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models/gemini-3.1-pro-preview",
      "sourceType": "official",
      "verifiedAt": "2026-08-18",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Model card verifies the Gemini 3.1 Pro Preview endpoint, multimodal inputs, text output, token limits, and documented capabilities."
    },
    {
      "id": "google-gemini-3-flash-preview-model",
      "title": "Gemini 3 Flash Preview",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models/gemini-3-flash-preview",
      "sourceType": "official",
      "verifiedAt": "2026-08-18",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Model card verifies the Gemini 3 Flash Preview endpoint, multimodal inputs, text output, token limits, and documented capabilities."
    },
    {
      "id": "google-gemini-3-1-flash-lite-model",
      "title": "Gemini 3.1 Flash-Lite",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models/gemini-3.1-flash-lite",
      "sourceType": "official",
      "verifiedAt": "2026-08-18",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Model card verifies the stable Gemini 3.1 Flash-Lite endpoint, multimodal inputs, text output, token limits, and documented capabilities."
    },
    {
      "id": "google-gemini-3-5-live-translate-model",
      "title": "Gemini 3.5 Live Translate",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models/gemini-3.5-live-translate-preview",
      "sourceType": "official",
      "verifiedAt": "2026-08-18",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Model card verifies the Gemini 3.5 Live Translate preview endpoint, audio input, translated-audio and transcript output, and token limits."
    },
    {
      "id": "google-live-api-guide",
      "title": "Gemini Live API",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/live-api",
      "sourceType": "official",
      "verifiedAt": "2026-08-18",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Live API guide verifies realtime audio model operation and the preview endpoint lifecycle applies to the live-model records."
    },
    {
      "id": "google-gemini-3-1-flash-live-model",
      "title": "Gemini 3.1 Flash Live Preview",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models/gemini-3.1-flash-live-preview",
      "sourceType": "official",
      "verifiedAt": "2026-08-18",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Model card verifies the Gemini 3.1 Flash Live preview endpoint, supported realtime multimodal inputs, text and audio output, token limits, and capabilities."
    },
    {
      "id": "google-gemini-3-1-flash-tts-model",
      "title": "Gemini 3.1 Flash TTS Preview",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models/gemini-3.1-flash-tts-preview",
      "sourceType": "official",
      "verifiedAt": "2026-08-18",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Model card verifies the Gemini 3.1 Flash TTS preview endpoint, text input, audio output, token limits, and audio generation capability."
    },
    {
      "id": "google-tts-guide",
      "title": "Text-to-speech generation",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/speech-generation",
      "sourceType": "official",
      "verifiedAt": "2026-08-18",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Text-to-Speech guide verifies controllable speech generation and streaming support for the Gemini 3.1 Flash TTS preview endpoint."
    },
    {
      "id": "google-gemini-3-1-flash-lite-image-model",
      "title": "Gemini 3.1 Flash Lite Image",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models/gemini-3.1-flash-lite-image",
      "sourceType": "official",
      "verifiedAt": "2026-08-18",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Model card verifies the stable Gemini 3.1 Flash Lite Image endpoint, text and image inputs, text and image output, token limits, and image generation and editing capabilities."
    },
    {
      "id": "google-gemini-3-pro-image-model",
      "title": "Gemini 3 Pro Image",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models/gemini-3-pro-image",
      "sourceType": "official",
      "verifiedAt": "2026-08-18",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Model card verifies the stable Gemini 3 Pro Image endpoint, text and image inputs and outputs, token limits, and image generation capability."
    },
    {
      "id": "google-gemini-omni-flash-model",
      "title": "Gemini Omni Flash",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models/gemini-omni-flash",
      "sourceType": "official",
      "verifiedAt": "2026-11-06",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Model card for gemini-omni-1.2-flash and gemini-omni-1.1-flash: text, image and video input, video output with native audio; 131,072 input and 65,536 output tokens; 1.2 makes 3 to 15 second clips, 1.1 makes 3 to 10 second clips, both at 360p to 4K and 24 FPS. gemini-omni-1.1-flash routes to 1.2 on 2026-12-03, and gemini-omni-flash-preview is shut down."
    },
    {
      "id": "google-omni-flash-guide",
      "title": "Generate and edit videos with Gemini Omni Flash",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/omni",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Gemini Omni Flash guide verifies preview lifecycle, conversational video editing, native multimodal context, and generated video with audio."
    },
    {
      "id": "google-gemini-deprecations",
      "title": "Gemini deprecations",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/deprecations",
      "sourceType": "official",
      "verifiedAt": "2026-08-18",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Lifecycle table verifies published stable or preview status and announced shutdown dates only for the Gemini endpoints that reference it."
    },
    {
      "id": "epoch-gpqa-diamond",
      "title": "GPQA Diamond benchmark (Epoch AI)",
      "publisher": "Epoch AI",
      "url": "https://epoch.ai/benchmarks/gpqa-diamond",
      "sourceType": "independent",
      "verifiedAt": "2026-10-12",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "198 four-option graduate-level questions in biology, physics and chemistry, written by domain experts; random guessing scores 25%. Epoch runs the evaluation itself."
    },
    {
      "id": "epoch-frontiermath",
      "title": "FrontierMath benchmark (Epoch AI)",
      "publisher": "Epoch AI",
      "url": "https://epoch.ai/benchmarks/frontiermath",
      "sourceType": "independent",
      "verifiedAt": "2026-10-12",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "Original research-level mathematics problems written and vetted by expert mathematicians. Tiers 1-3 v2 (released 2026-06-12) is the 295-problem base set after a revision that corrected errors in 42% of problems. Scores come from Epoch-run private sets. OpenAI funded the benchmark and has access to part of it."
    },
    {
      "id": "epoch-simpleqa-verified",
      "title": "SimpleQA Verified benchmark (Epoch AI)",
      "publisher": "Epoch AI",
      "url": "https://epoch.ai/benchmarks/simpleqa-verified",
      "sourceType": "independent",
      "verifiedAt": "2026-10-12",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "A 1,000-question factoid benchmark from Google, building on SimpleQA from OpenAI. Epoch runs it itself and reports the proportion answered correctly, not the F1-based score Google reports."
    },
    {
      "id": "deepseek-v4-1-flash-release",
      "title": "DeepSeek-V4.1-Flash release",
      "publisher": "DeepSeek",
      "url": "https://api-docs.deepseek.com/news/news260910",
      "sourceType": "official",
      "verifiedAt": "2026-10-12",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official release note, 2026-09-10: 552B-parameter MoE with native visual understanding; 8B active parameters for input and 16B for output."
    },
    {
      "id": "deepseek-v4-1-flash-weights",
      "title": "DeepSeek-V4.1-Flash model weights",
      "publisher": "DeepSeek",
      "url": "https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash",
      "sourceType": "official",
      "verifiedAt": "2026-10-12",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official open-weights repository, created 2026-09-10, MIT licence, image-text-to-text."
    },
    {
      "id": "mistral-large-4-release",
      "title": "Introducing Mistral Large 4",
      "publisher": "Mistral AI",
      "url": "https://mistral.ai/news/mistral-large-4/",
      "sourceType": "official",
      "verifiedAt": "2026-10-12",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Public preview announced 2026-10-06 at $1.36 / $4.18 per 1M input / output tokens; weights promised by the end of October."
    },
    {
      "id": "mistral-large-4-model",
      "title": "Mistral Large 4 model card",
      "publisher": "Mistral AI",
      "url": "https://docs.mistral.ai/models/mistral-large-4-0",
      "sourceType": "official",
      "verifiedAt": "2026-11-03",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Model id mistral-large-4-0, still labelled public preview, 1M context, 1.05T total and 52B active parameters, 1.6B vision encoder; $0.70 input, $0.07 cached input and $2.10 output per 1M tokens."
    },
    {
      "id": "qwen-3-8-omni-flash-model",
      "title": "qwen3.8-omni-flash model information",
      "publisher": "Alibaba Cloud",
      "url": "https://www.alibabacloud.com/help/en/model-studio/qwen3-8-omni-flash",
      "sourceType": "official",
      "verifiedAt": "2026-10-12",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Text, image, audio and video input; text output; 1M-token context; 131,072 max output tokens."
    },
    {
      "id": "qwen-3-8-omni-flash-blog",
      "title": "Qwen3.8-Omni-Flash: Omni Senses. Agentic Delivery.",
      "publisher": "Alibaba Cloud",
      "url": "https://www.alibabacloud.com/blog/qwen3-8-omni-flash-omni-senses--agentic-delivery-_603580",
      "sourceType": "official",
      "verifiedAt": "2026-10-12",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official announcement published 2026-09-20."
    },
    {
      "id": "alibaba-model-studio-pricing",
      "title": "Alibaba Cloud Model Studio model pricing",
      "publisher": "Alibaba Cloud",
      "url": "https://www.alibabacloud.com/help/en/model-studio/model-pricing",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "pricing"
      ],
      "reuseNote": "International (Singapore) rates used."
    },
    {
      "id": "google-nano-banana-2-1-model",
      "title": "Gemini Nano Banana 2.1",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models/gemini-nano-banana-2.1",
      "sourceType": "official",
      "verifiedAt": "2026-10-12",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Stable model gemini-nano-banana-2.1: text, image, video and PDF input; image and text output; 1K, 2K and 4K output."
    },
    {
      "id": "google-gemini-3-8-live-release",
      "title": "Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking",
      "publisher": "Google",
      "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-live-gemini-3-8-live-extended-thinking/",
      "sourceType": "official",
      "verifiedAt": "2026-10-12",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Two speech-to-speech models for the Gemini API and Google AI Studio."
    },
    {
      "id": "google-gemini-3-8-tts-release",
      "title": "Gemini 3.8 text-to-speech models",
      "publisher": "Google",
      "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-text-to-speech/",
      "sourceType": "official",
      "verifiedAt": "2026-10-12",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Gemini 3.8 Flash TTS and Flash-Lite TTS announcement."
    },
    {
      "id": "openai-gpt-live-1-model",
      "title": "GPT-Live 1 model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-live-1",
      "sourceType": "official",
      "verifiedAt": "2026-10-12",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Full-duplex voice model; audio and text in and out; $0.05 per voice-session minute billed per second; reasoning and tools delegated to a separately billed backend model."
    },
    {
      "id": "openai-api-changelog",
      "title": "OpenAI API changelog",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/changelog",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Release dates: gpt-image-2.5-sunburst and gpt-image-2.5-flare 2026-09-08; gpt-live-1 generally available 2026-09-10."
    },
    {
      "id": "openai-gpt-image-2-5-sunburst-model",
      "title": "GPT Image 2.5 Sunburst model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-image-2.5-sunburst",
      "sourceType": "official",
      "verifiedAt": "2026-10-12",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Most capable image generation and editing model; token rates match GPT Image 2."
    },
    {
      "id": "openai-gpt-image-2-5-flare-model",
      "title": "GPT Image 2.5 Flare model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-image-2.5-flare",
      "sourceType": "official",
      "verifiedAt": "2026-10-12",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Fastest model for everyday image generation; token rates match GPT Image 2."
    },
    {
      "id": "xai-grok-imagine-video-1-5-lite-model",
      "title": "grok-imagine-video-1.5-lite",
      "publisher": "xAI",
      "url": "https://docs.x.ai/developers/models/grok-imagine-video-1.5-lite",
      "sourceType": "official",
      "verifiedAt": "2026-10-12",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Text and image to video at $0.020 per output second; batch API supported."
    },
    {
      "id": "openai-gpt-5-6-sol-model",
      "title": "GPT-5.6 Sol model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-5.6-sol",
      "sourceType": "official",
      "verifiedAt": "2026-10-12",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "1.05M context, 128K max output, knowledge cutoff 2026-02-16, six reasoning-effort levels."
    },
    {
      "id": "openai-gpt-5-6-terra-model",
      "title": "GPT-5.6 Terra model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-5.6-terra",
      "sourceType": "official",
      "verifiedAt": "2026-10-12",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "1.05M context, 128K max output, knowledge cutoff 2026-02-16, six reasoning-effort levels."
    },
    {
      "id": "openai-gpt-5-6-luna-model",
      "title": "GPT-5.6 Luna model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-5.6-luna",
      "sourceType": "official",
      "verifiedAt": "2026-10-12",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "1.05M context, 128K max output, knowledge cutoff 2026-02-16, six reasoning-effort levels."
    },
    {
      "id": "arena-text-to-image",
      "title": "Arena Text-to-Image leaderboard",
      "publisher": "Arena",
      "url": "https://arena.ai/leaderboard/image/text-to-image",
      "sourceType": "independent",
      "verifiedAt": "2026-11-03",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "People compare two anonymous images generated from the same prompt and vote for the better one. Arena fits a Bradley-Terry model to the votes and reports ratings on an Elo-like scale."
    },
    {
      "id": "arena-image-edit",
      "title": "Arena Image Edit leaderboard",
      "publisher": "Arena",
      "url": "https://arena.ai/leaderboard/image/image-edit",
      "sourceType": "independent",
      "verifiedAt": "2026-11-03",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "People compare two anonymous edits of the same image from the same instruction and vote for the better one. Arena fits a Bradley-Terry model to the votes and reports ratings on an Elo-like scale."
    },
    {
      "id": "arena-text-to-video",
      "title": "Arena Text-to-Video leaderboard",
      "publisher": "Arena",
      "url": "https://arena.ai/leaderboard/video/text-to-video",
      "sourceType": "independent",
      "verifiedAt": "2026-11-03",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "People compare two anonymous videos generated from the same prompt and vote for the better one. Arena fits a Bradley-Terry model to the votes and reports ratings on an Elo-like scale."
    },
    {
      "id": "arena-image-to-video",
      "title": "Arena Image-to-Video leaderboard",
      "publisher": "Arena",
      "url": "https://arena.ai/leaderboard/video/image-to-video",
      "sourceType": "independent",
      "verifiedAt": "2026-11-03",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "People compare two anonymous videos animated from the same image and prompt and vote for the better one. Arena fits a Bradley-Terry model to the votes and reports ratings on an Elo-like scale."
    },
    {
      "id": "arena-text",
      "title": "Arena Text leaderboard",
      "publisher": "Arena",
      "url": "https://arena.ai/leaderboard/chat/text",
      "sourceType": "independent",
      "verifiedAt": "2026-11-03",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "People put the same prompt to two anonymous models and vote for the better answer. Arena fits a Bradley-Terry model to the votes; the rating used here is the style-controlled one, which is the default view on arena.ai."
    },
    {
      "id": "elevenlabs-api-pricing",
      "title": "ElevenLabs API pricing",
      "publisher": "ElevenLabs",
      "url": "https://elevenlabs.io/pricing/api",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "xai-developer-models",
      "title": "xAI models and pricing",
      "publisher": "xAI",
      "url": "https://docs.x.ai/developers/models",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "alibaba-model-studio-new-models",
      "title": "Alibaba Cloud Model Studio newly released models",
      "publisher": "Alibaba Cloud",
      "url": "https://www.alibabacloud.com/help/en/model-studio/newly-released-models",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "zai-pricing",
      "title": "Z.ai API pricing",
      "publisher": "Z.ai",
      "url": "https://docs.z.ai/guides/overview/pricing",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "zai-release-notes",
      "title": "Z.ai release notes",
      "publisher": "Z.ai",
      "url": "https://docs.z.ai/release-notes/new-released",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "xiaomi-mimo-pricing",
      "title": "Xiaomi MiMo API pay-as-you-go pricing",
      "publisher": "Xiaomi",
      "url": "https://platform.xiaomimimo.com/docs/en-US/price/pay-as-you-go",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "xiaomi-mimo-model-updates",
      "title": "Xiaomi MiMo model updates",
      "publisher": "Xiaomi",
      "url": "https://platform.xiaomimimo.com/docs/en-US/updates/model",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "google-gemini-4-argon-release",
      "title": "Gemini 4 Argon announcement",
      "publisher": "Google",
      "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "meta-model-api-pricing",
      "title": "Meta Model API pricing and rate limits",
      "publisher": "Meta",
      "url": "https://dev.meta.ai/docs/pricing-rate-limits",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "meta-muse-wonder-model-card",
      "title": "Muse Wonder model card",
      "publisher": "Meta",
      "url": "https://research.meta.ai/static/muse-wonder-model-card",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "meta-muse-glimmer-docs",
      "title": "Muse Glimmer documentation",
      "publisher": "Meta",
      "url": "https://dev.meta.ai/docs/muse-glimmer",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "meta-muse-glimmer-release",
      "title": "Introducing Muse Glimmer",
      "publisher": "Meta",
      "url": "https://research.meta.ai/blog/introducing-muse-glimmer-open-agentic-model",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "zai-glm-5-3-docs",
      "title": "GLM-5.3 documentation",
      "publisher": "Z.ai",
      "url": "https://docs.z.ai/guides/llm/glm-5.3",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "zai-glm-5-3-flash-docs",
      "title": "GLM-5.3-Flash documentation",
      "publisher": "Z.ai",
      "url": "https://docs.z.ai/guides/vlm/glm-5.3-flash",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "deepseek-v4-pro-0813-release",
      "title": "DeepSeek-V4-Pro update (2026-08-13)",
      "publisher": "DeepSeek",
      "url": "https://api-docs.deepseek.com/news/news260813",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "deepseek-v4-pro-0813-weights",
      "title": "DeepSeek-V4-Pro-0813 model weights",
      "publisher": "DeepSeek",
      "url": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro-0813",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "mimo-v2-6-pro-weights",
      "title": "MiMo-V2.6-Pro model weights",
      "publisher": "Xiaomi",
      "url": "https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Pro-RL",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "mimo-v2-6-flash-weights",
      "title": "MiMo-V2.6-Flash model weights",
      "publisher": "Xiaomi",
      "url": "https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "qwen-3-8-flash-next-weights",
      "title": "Qwen3.8-Flash-Next model card",
      "publisher": "Alibaba Qwen",
      "url": "https://huggingface.co/Qwen/Qwen3.8-Flash-Next",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "nvidia-nemotron-3-5-lightning-weights",
      "title": "NVIDIA Nemotron 3.5 Lightning 30B-A3B model card",
      "publisher": "NVIDIA",
      "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "nvidia-nemotron-3-5-lightning-release",
      "title": "NVIDIA Nemotron 3.5 Lightning announcement",
      "publisher": "NVIDIA",
      "url": "https://developer.nvidia.com/blog/nvidia-nemotron-3-5-lightning-delivers-fast-accurate-specialized-task-execution-for-long-running-agents/",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "ibm-granite-4-2-30b-weights",
      "title": "Granite 4.2 30B model card",
      "publisher": "IBM",
      "url": "https://huggingface.co/ibm-granite/granite-4.2-30b",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "ibm-granite-4-2-docs",
      "title": "Granite 4.2 documentation",
      "publisher": "IBM",
      "url": "https://www.ibm.com/granite/docs/models/granite4-2",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "inception-models",
      "title": "Inception models",
      "publisher": "Inception",
      "url": "https://docs.inceptionlabs.ai/get-started/models",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "inception-mercury-2-5-release",
      "title": "Introducing Mercury 2.5",
      "publisher": "Inception",
      "url": "https://www.inceptionlabs.ai/blog/introducing-mercury-2-5",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "upstage-solar-pro-4-docs",
      "title": "Solar Pro 4 documentation",
      "publisher": "Upstage",
      "url": "https://console.upstage.ai/docs/models/solar-pro-4",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "stepfun-step-5-preview-docs",
      "title": "Step 5 Preview documentation",
      "publisher": "StepFun",
      "url": "https://platform.stepfun.ai/docs/en/guides/models/step-5-preview",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "stepfun-pricing",
      "title": "StepFun API pricing",
      "publisher": "StepFun",
      "url": "https://platform.stepfun.ai/docs/en/guides/pricing/details",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "openai-gpt-realtime-2-model",
      "title": "GPT-Realtime-2 model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-realtime-2",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "openai-gpt-transcribe-model",
      "title": "GPT Transcribe model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-transcribe",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "elevenlabs-eleven-v4-release",
      "title": "Eleven v4 announcement",
      "publisher": "ElevenLabs",
      "url": "https://elevenlabs.io/blog/eleven-v4",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "xai-grok-voice-transcribe-2-model",
      "title": "Grok Voice Transcribe 2.0 model",
      "publisher": "xAI",
      "url": "https://docs.x.ai/developers/models/grok-voice-transcribe-2.0",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "xai-grok-voice-transcribe-2-release",
      "title": "Grok Voice Transcribe 2.0 announcement",
      "publisher": "xAI",
      "url": "https://x.ai/news/grok-voice-transcribe-2",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "microsoft-mai-transcribe-docs",
      "title": "MAI-Transcribe in Azure Speech",
      "publisher": "Microsoft",
      "url": "https://learn.microsoft.com/en-us/azure/ai-services/speech-service/mai-transcribe",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "microsoft-mai-transcribe-2-release",
      "title": "MAI-Transcribe-2 announcement",
      "publisher": "Microsoft AI",
      "url": "https://microsoft.ai/news/mai-transcribe-2",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "cartesia-sonic-models",
      "title": "Cartesia TTS models",
      "publisher": "Cartesia",
      "url": "https://docs.cartesia.ai/build-with-cartesia/tts-models/latest",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "cartesia-pricing",
      "title": "Cartesia pricing",
      "publisher": "Cartesia",
      "url": "https://cartesia.ai/pricing",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "google-gemini-3-5-transcribe-model",
      "title": "Gemini 3.5 Transcribe model",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models/gemini-3.5-transcribe",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "xai-grok-imagine-video-1-5-model",
      "title": "Grok Imagine Video 1.5 model",
      "publisher": "xAI",
      "url": "https://docs.x.ai/developers/models/grok-imagine-video-1.5",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "google-veo-3-1-lite-model",
      "title": "Veo 3.1 Lite Preview model",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/models/veo-3.1-lite-generate-preview",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "alibaba-wan-3-video-api",
      "title": "Wan3.0 video generation API reference",
      "publisher": "Alibaba Cloud",
      "url": "https://www.alibabacloud.com/help/en/model-studio/wan3-video-generation-api-reference",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "byteplus-modelark-releases",
      "title": "BytePlus ModelArk model releases",
      "publisher": "BytePlus",
      "url": "https://docs.byteplus.com/en/docs/ModelArk/model-release-announcement",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "byteplus-modelark-pricing",
      "title": "BytePlus ModelArk pricing",
      "publisher": "BytePlus",
      "url": "https://docs.byteplus.com/en/docs/ModelArk/1099320",
      "sourceType": "official",
      "verifiedAt": "2026-10-13",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "artificial-analysis-speech-arena",
      "title": "Artificial Analysis Speech Arena leaderboard (text to speech)",
      "publisher": "Artificial Analysis",
      "url": "https://artificialanalysis.ai/text-to-speech/leaderboard",
      "sourceType": "independent",
      "verifiedAt": "2026-11-03",
      "datasetVersion": "Provider voice leaderboard, read 2026-11-03",
      "license": "Artificial Analysis Terms of Use",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "Listeners compare two anonymous clips of the same text and vote for the more natural one; Artificial Analysis fits Elo ratings to the votes, using each provider's own voices. Cited with a link as Artificial Analysis asks; its terms reserve redistribution, so these scores are shown on the page and left out of the dataset download."
    },
    {
      "id": "artificial-analysis-aa-wer",
      "title": "Artificial Analysis speech-to-text leaderboard (AA-WER v2, non-streaming)",
      "publisher": "Artificial Analysis",
      "url": "https://artificialanalysis.ai/speech-to-text/non-streaming",
      "sourceType": "independent",
      "verifiedAt": "2026-11-03",
      "datasetVersion": "AA-WER v2, read 2026-11-03",
      "license": "Artificial Analysis Terms of Use",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "Artificial Analysis transcribes the same held-out audio through each provider API and reports the weighted word error rate. Cited with a link as Artificial Analysis asks; its terms reserve redistribution, so these scores are shown on the page and left out of the dataset download."
    },
    {
      "id": "artificial-analysis-aa-wer-streaming",
      "title": "Artificial Analysis speech-to-text leaderboard (AA-WER Streaming)",
      "publisher": "Artificial Analysis",
      "url": "https://artificialanalysis.ai/speech-to-text/streaming",
      "sourceType": "independent",
      "verifiedAt": "2026-11-03",
      "datasetVersion": "AA-WER Streaming, read 2026-11-03",
      "license": "Artificial Analysis Terms of Use",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "Artificial Analysis streams the same held-out audio through each provider realtime API and reports the word error rate of the final transcript. Cited with a link as Artificial Analysis asks; its terms reserve redistribution, so these scores are shown on the page and left out of the dataset download."
    },
    {
      "id": "artificial-analysis-video-arena-t2v",
      "title": "Artificial Analysis Video Arena leaderboard (AA-Video-T2V v2.1)",
      "publisher": "Artificial Analysis",
      "url": "https://artificialanalysis.ai/video/leaderboard/text-to-video",
      "sourceType": "independent",
      "verifiedAt": "2026-11-03",
      "datasetVersion": "AA-Video-T2V v2.1, read 2026-11-03",
      "license": "Artificial Analysis Terms of Use",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "People compare two anonymous videos generated from the same prompt and vote for the better one; Artificial Analysis fits Elo ratings to the votes. Cited with a link as Artificial Analysis asks; its terms reserve redistribution, so these scores are shown on the page and left out of the dataset download."
    },
    {
      "id": "artificial-analysis-video-arena-i2v",
      "title": "Artificial Analysis Video Arena leaderboard (AA-Video-I2V v2.1)",
      "publisher": "Artificial Analysis",
      "url": "https://artificialanalysis.ai/video/leaderboard/image-to-video",
      "sourceType": "independent",
      "verifiedAt": "2026-11-03",
      "datasetVersion": "AA-Video-I2V v2.1, read 2026-11-03",
      "license": "Artificial Analysis Terms of Use",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "People compare two anonymous videos animated from the same image and prompt and vote for the better one; Artificial Analysis fits Elo ratings to the votes. Cited with a link as Artificial Analysis asks; its terms reserve redistribution, so these scores are shown on the page and left out of the dataset download."
    },
    {
      "id": "meta-muse-spark-open-release",
      "title": "Introducing Muse Spark Open",
      "publisher": "Meta",
      "url": "https://research.meta.ai/blog/muse-spark-open",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official announcement, 2026-10-15: 816B total and 42B active parameters, 1M-token context, weights under the Muse Model License."
    },
    {
      "id": "meta-muse-spark-open-weights",
      "title": "Muse Spark Open model card",
      "publisher": "Meta",
      "url": "https://huggingface.co/meta-models/Muse-Spark-Open-816B-A42B",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official open-weights repository: text, image, video and audio in, text out; no first-party API."
    },
    {
      "id": "nvidia-nemotron-3-5-super-vl-release",
      "title": "NVIDIA expands Nemotron with Super VL and Bolt models",
      "publisher": "NVIDIA",
      "url": "https://blogs.nvidia.com/blog/nemotron-3-5-super-vl/",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "nvidia-nemotron-3-5-super-vl-weights",
      "title": "NVIDIA Nemotron 3.5 Super VL 120B-A12B model card",
      "publisher": "NVIDIA",
      "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "nvidia-nemotron-3-5-bolt-weights",
      "title": "NVIDIA Nemotron 3.5 Bolt 74B-A9B 24GB model card",
      "publisher": "NVIDIA",
      "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GB-GGUF",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "reflection-beam-release",
      "title": "Beam is now available",
      "publisher": "Reflection AI",
      "url": "https://reflection.ai/blog/beam-is-available",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "reflection-beam-weights",
      "title": "Beam 1.0 501B-A23B model card",
      "publisher": "Reflection AI",
      "url": "https://huggingface.co/reflection/Beam-1.0-501B-A23B",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "reflection-developer-docs",
      "title": "Reflection developer documentation",
      "publisher": "Reflection AI",
      "url": "https://developers.reflection.ai/introduction",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "minimax-m3-1-flash-release",
      "title": "MiniMax M3.1 Flash",
      "publisher": "MiniMax",
      "url": "https://www.minimax.io/blog/minimax-m31f",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "minimax-m3-1-flash-weights",
      "title": "MiniMax M3.1 Flash model card",
      "publisher": "MiniMax",
      "url": "https://huggingface.co/MiniMaxAI/MiniMax-M3.1-Flash",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "minimax-text-generation-docs",
      "title": "MiniMax API text generation guide",
      "publisher": "MiniMax",
      "url": "https://platform.minimax.io/docs/guides/text-generation",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "inception-mercury-3-release",
      "title": "Diffusion at Scale: Introducing Mercury 3",
      "publisher": "Inception",
      "url": "https://www.inceptionlabs.ai/blog/introducing-mercury-3",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "xai-grok-voice-3-release",
      "title": "Introducing Grok Voice 3.0",
      "publisher": "xAI",
      "url": "https://x.ai/news/grok-voice-3",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "xai-grok-voice-3-model",
      "title": "Grok Voice 3.0 model",
      "publisher": "xAI",
      "url": "https://docs.x.ai/developers/models/grok-voice-3.0",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "mistral-voxtral-3-release",
      "title": "Voxtral 3",
      "publisher": "Mistral AI",
      "url": "https://mistral.ai/news/voxtral-3/",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "mistral-voxtral-small-transcribe-3-model",
      "title": "Voxtral Small Transcribe 3",
      "publisher": "Mistral AI",
      "url": "https://docs.mistral.ai/models/voxtral-small-transcribe-3-0",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "mistral-voxtral-mini-tts-3-model",
      "title": "Voxtral TTS 3",
      "publisher": "Mistral AI",
      "url": "https://docs.mistral.ai/models/voxtral-mini-tts-3-0",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "google-gemini-4-argon-ga",
      "title": "Gemini 4 Argon is now available for everyone",
      "publisher": "Google",
      "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon/",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "deepseek-v4-1-flash-1017-release",
      "title": "DeepSeek-V4.1-Flash-1017",
      "publisher": "DeepSeek",
      "url": "https://api-docs.deepseek.com/news/news261017/",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official release note, 2026-10-17: a minor update served under the same API name at unchanged prices."
    },
    {
      "id": "deepseek-v4-1-flash-1017-weights",
      "title": "DeepSeek-V4.1-Flash-1017 model weights",
      "publisher": "DeepSeek",
      "url": "https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017",
      "sourceType": "official",
      "verifiedAt": "2026-10-23",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official open-weights repository, MIT licence, image-text-to-text, 1M-token context."
    },
    {
      "id": "openai-gpt-6-1-terra-model",
      "title": "GPT-6.1 Terra model reference",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-6.1-terra",
      "sourceType": "official",
      "verifiedAt": "2026-10-28",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "kimi-k3-1-pro-release",
      "title": "Kimi K3.1 Pro",
      "publisher": "Moonshot AI",
      "url": "https://www.kimi.ai/blog/kimi-k3-1-pro",
      "sourceType": "official",
      "verifiedAt": "2026-10-28",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "kimi-k3-1-pro-quickstart",
      "title": "Kimi K3.1 Pro quickstart",
      "publisher": "Moonshot AI",
      "url": "https://platform.kimi.ai/docs/guide/kimi-k3-1-pro-quickstart",
      "sourceType": "official",
      "verifiedAt": "2026-11-05",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "elorian-glimpse-release",
      "title": "Introducing Glimpse: visual thinking that stays grounded",
      "publisher": "Elorian",
      "url": "https://elorian.ai/blog/introducing-glimpse",
      "sourceType": "official",
      "verifiedAt": "2026-10-29",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "elorian-models",
      "title": "Elorian models",
      "publisher": "Elorian",
      "url": "https://elorian.ai/models.html",
      "sourceType": "official",
      "verifiedAt": "2026-10-29",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "microsoft-mai-streaming-voice-release",
      "title": "Our first streaming transcription model debuts at no. 1 on Artificial Analysis",
      "publisher": "Microsoft AI",
      "url": "https://microsoft.ai/news/our-first-streaming-transcription-model/",
      "sourceType": "official",
      "verifiedAt": "2026-10-29",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "microsoft-foundry-mai-voice-2-1",
      "title": "MAI-Voice-2.1 in the Microsoft Foundry model catalog",
      "publisher": "Microsoft",
      "url": "https://ai.azure.com/catalog/models/MAI-Voice-2.1",
      "sourceType": "official",
      "verifiedAt": "2026-10-29",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "microsoft-foundry-mai-voice-2-1-flash",
      "title": "MAI-Voice-2.1-Flash in the Microsoft Foundry model catalog",
      "publisher": "Microsoft",
      "url": "https://ai.azure.com/catalog/models/MAI-Voice-2.1-Flash",
      "sourceType": "official",
      "verifiedAt": "2026-10-29",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "microsoft-foundry-mai-transcribe-2-streaming",
      "title": "MAI-Transcribe-2-Streaming in the Microsoft Foundry model catalog",
      "publisher": "Microsoft",
      "url": "https://ai.azure.com/catalog/models/MAI-Transcribe-2-Streaming",
      "sourceType": "official",
      "verifiedAt": "2026-10-29",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing."
    },
    {
      "id": "meituan-longcat-changelog",
      "title": "LongCat API Platform change log",
      "publisher": "Meituan",
      "url": "https://longcat.chat/platform/docs/change-log",
      "sourceType": "official",
      "verifiedAt": "2026-11-03",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Version 2026-11-02 makes LongCat-2.5 available on the LongCat API Platform: native multimodality with agentic coding, long-horizon agents for complex codebases, and compatibility with Claude Code, Codex and other agent harnesses. LongCat-2.5-Preview had been available since 2026-09-25."
    },
    {
      "id": "meituan-longcat-model-params",
      "title": "LongCat API model parameters",
      "publisher": "Meituan",
      "url": "https://longcat.chat/platform/docs/api/model-params",
      "sourceType": "official",
      "verifiedAt": "2026-11-03",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "LongCat-2.5: a 1,048,576-token context window counting input and output, max_tokens up to 262,144, and a thinking budget of up to 262,144 tokens."
    },
    {
      "id": "meituan-longcat-visual-understanding",
      "title": "LongCat API visual understanding",
      "publisher": "Meituan",
      "url": "https://longcat.chat/platform/docs/image-video-understanding",
      "sourceType": "official",
      "verifiedAt": "2026-11-03",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "LongCat-2.5 accepts images (up to 100 a request, 10 MB each) and videos by URL (up to 10 a request, 100 MB and under an hour each) alongside text."
    },
    {
      "id": "meituan-longcat-2-5-pricing",
      "title": "LongCat-2.5 pricing",
      "publisher": "Meituan",
      "url": "https://longcat.chat/platform/docs/pricing/longcat-2.5",
      "sourceType": "official",
      "verifiedAt": "2026-11-03",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "pricing"
      ],
      "reuseNote": "Pay-as-you-go LongCat-2.5 at a discounted, limited-time price: $0.30 uncached input, $0.006 cached input and $1.20 output per 1M tokens. The page gives no list price and no end date."
    },
    {
      "id": "assemblyai-tts-1-release",
      "title": "Introducing AssemblyAI-TTS-1: text-to-speech built for voice agents",
      "publisher": "AssemblyAI",
      "url": "https://www.assemblyai.com/blog/text-to-speech-api",
      "sourceType": "official",
      "verifiedAt": "2026-11-03",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Launch post: AssemblyAI-TTS-1 is generally available at a list price of $15 per 1M characters, with every voice, language, instant cloning, word timestamps and US or EU processing included; 4,000+ voices; English, Spanish, French, German, Italian and Portuguese at launch; 90 ms to first audio in production. Its comparisons with other models are AssemblyAI tests (eval dated 2026-10-28)."
    },
    {
      "id": "assemblyai-models",
      "title": "AssemblyAI models",
      "publisher": "AssemblyAI",
      "url": "https://www.assemblyai.com/docs/getting-started/models",
      "sourceType": "official",
      "verifiedAt": "2026-11-03",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Model list: AssemblyAI-TTS-1 at $15 per 1M input characters, billed per character of input text, streamed over WebSocket, HTTP and an OpenAI-compatible route; every voice speaks all six languages."
    },
    {
      "id": "mistral-large-4-open-weights",
      "title": "Mistral Large 4 (Le Chonk) is now open weight",
      "publisher": "Mistral AI",
      "url": "https://mistral.ai/news/mistral-large-4-oss",
      "sourceType": "official",
      "verifiedAt": "2026-11-03",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Announcement of 2026-10-29: weights of the improved Mistral Large 4 released under the Mistral Ecosystem Open-Weight License, a modified MIT licence under which businesses selling it as a hosted model service need a commercial licence; API price $0.70 input, $0.07 cached input and $2.10 output per 1M tokens from that day."
    },
    {
      "id": "mistral-large-4-weights",
      "title": "Mistral Large 4 model weights",
      "publisher": "Mistral AI",
      "url": "https://huggingface.co/mistralai/Mistral-Large-4-1T-A52B",
      "sourceType": "official",
      "verifiedAt": "2026-11-03",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Official open-weights repository (FP8): about 1T parameters, 49B active per token (52B with embeddings and output layers), 1M context, text and image input; an NVFP4 checkpoint is published separately."
    },
    {
      "id": "google-gemini-changelog-2026-11-02",
      "title": "Gemini API release notes: November 2, 2026",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/changelog#11-02-2026",
      "sourceType": "official",
      "verifiedAt": "2026-11-03",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Release note verifies that gemini-2.5-pro was deprecated on 2026-11-02 and is no longer offered in Google AI Studio, and that from 2026-12-01 API requests to it are routed to gemini-3.8-flash and billed at Gemini 3.8 Flash rates."
    },
    {
      "id": "google-gemini-3-preview-shutdowns",
      "title": "Gemini deprecations: Gemini 3 preview models",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/deprecations#gemini-3-preview-models",
      "sourceType": "official",
      "verifiedAt": "2026-11-03",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "identity"
      ],
      "reuseNote": "Lists gemini-3.1-flash-tts-preview with a shutdown date of 2026-11-17 and gemini-3.8-flash-tts or gemini-3.8-flash-lite-tts as its replacements."
    },
    {
      "id": "epoch-benchmark-data-2026-11-03",
      "title": "Epoch AI Benchmarking Hub data (benchmark_data.zip)",
      "publisher": "Epoch AI",
      "url": "https://epoch.ai/benchmarks/use-this-data",
      "sourceType": "independent",
      "verifiedAt": "2026-11-03",
      "datasetVersion": "benchmark_data.zip retrieved 2026-11-03, sha256 prefix 333c5aa3e36a4ea1",
      "license": "CC-BY-4.0",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "Results Epoch AI ran itself (the GPQA Diamond, FrontierMath Tiers 1-3 v2 and SimpleQA Verified files), not the external files it collects from other publishers. A model maps only when its Epoch row is unambiguous: exact name, or a dated version pinned by the record release date. Where Epoch lists several reasoning-effort settings, the best published configuration is used and its row is recorded in sourceRow. Scores are converted from 0-1 to 0-100 and rounded to one decimal. Epoch ran the newest rows up to 2026-11-02; every result is dated to this capture. Credit: Epoch AI, Capabilities and benchmarking, epoch.ai, CC BY 4.0."
    },
    {
      "id": "arena-leaderboard-dataset-2026-11-03",
      "title": "Arena leaderboard dataset (lmarena-ai/leaderboard-dataset)",
      "publisher": "Arena",
      "url": "https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset/tree/38eef63f8920f9e8a77d903928321ecd74ab8126",
      "sourceType": "independent",
      "verifiedAt": "2026-11-03",
      "datasetVersion": "lmarena-ai/leaderboard-dataset@38eef63f8920f9e8a77d903928321ecd74ab8126",
      "license": "CC-BY-4.0",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "Ratings from blind pairwise human votes, published by Arena (formerly LMArena) in its public leaderboard dataset and pinned here to one commit. The overall category of the latest split is used; for text, the style-controlled leaderboard, which is the default view on arena.ai, and for code, the WebDev leaderboard of Code Arena. A model maps only when its Arena row names the same model; where Arena lists several quality or resolution settings of one model, the best rated row is used and recorded in sourceRow. Ratings are rounded to whole points. Credit: Arena, lmarena-ai/leaderboard-dataset on Hugging Face, CC BY 4.0."
    },
    {
      "id": "arena-webdev",
      "title": "Arena WebDev leaderboard (Code Arena)",
      "publisher": "Arena",
      "url": "https://arena.ai/leaderboard/webdev",
      "sourceType": "independent",
      "verifiedAt": "2026-11-03",
      "datasetVersion": null,
      "license": "CC-BY-4.0",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "People give the same web development request to two anonymous models, try the apps they build and vote for the better one. Arena fits a Bradley-Terry model to the votes and reports ratings on an Elo-like scale."
    },
    {
      "id": "tencent-hy4-release",
      "title": "Introducing Hy4",
      "publisher": "Tencent",
      "url": "https://hy.tencent.ai/research/hy4?langVersion=en",
      "sourceType": "official",
      "verifiedAt": "2026-11-05",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Tencent released Hy4 on 2026-11-04 on Tencent Cloud and OpenRouter and in CodeBuddy, WorkBuddy, Yuanbao and ima. It adds image understanding and computer use across command-line and graphical interfaces to Hy4 preview, with scaled-up pre-training and post-training. API price per 1M tokens: 0.042 USD cached input, 0.834 USD input, 2.501 USD output, the same as Hy4 preview. Weights and a system card are promised for later. Benchmark figures on the page are Tencent own evaluations."
    },
    {
      "id": "tencent-hy4-preview-card",
      "title": "Hy4 preview model card",
      "publisher": "Tencent",
      "url": "https://huggingface.co/tencent/Hy4-preview",
      "sourceType": "official",
      "verifiedAt": "2026-11-05",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Open weights of Hy4 preview, released 2026-08-28 under the Apache License 2.0: a mixture-of-experts model with 770B total and 49B activated parameters and a 1M-token context. Cited for the predecessor of Hy4 only."
    },
    {
      "id": "moonshot-kimi-k3-1-pro-weights",
      "title": "Kimi K3.1 Pro model card and weights",
      "publisher": "Moonshot AI",
      "url": "https://huggingface.co/moonshotai/Kimi-K3.1-Pro",
      "sourceType": "official",
      "verifiedAt": "2026-11-05",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Open weights of Kimi K3.1 Pro: a mixture-of-experts model with 2.8T total and 104B activated parameters, a 1,048,576-token context and the MoonViT-V2 vision encoder, in MXFP4 weights from quantization-aware training (about 1.56 TB in 96 files). Released under the Kimi K3.1 Pro License, MIT-style with two conditions: a model-as-a-service business above 20M USD revenue over 12 months needs a separate agreement, and products above 100M monthly active users or 20M USD monthly revenue must display the model name."
    },
    {
      "id": "google-gemini-4-carbon-release",
      "title": "Introducing Gemini 4 Carbon",
      "publisher": "Google",
      "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-carbon/",
      "sourceType": "official",
      "verifiedAt": "2026-11-05",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing",
        "benchmark"
      ],
      "reuseNote": "Google announced Gemini 4 Carbon on 2026-10-28 and makes it available first to a subset of cyber defenders, then to more trusted partners through the Fairwind Program, before developers, businesses and consumers. Introductory price: 2 USD input and 10 USD output per 1M tokens, cached input 95% off; after the introductory period, 4 USD and 20 USD. The launch table compares Carbon with Gemini 4 Argon, GPT-6 Astra, Claude Opus 5.5 and Claude Fable 5.1, evaluated without guardrails."
    },
    {
      "id": "google-fairwind-program",
      "title": "Fairwind Program",
      "publisher": "Google DeepMind",
      "url": "https://deepmind.google/fairwind-program/",
      "sourceType": "official",
      "verifiedAt": "2026-11-05",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "The Fairwind Program gives high-priority cyber defenders such as governments, healthcare providers and telecommunications services early access to Google frontier models under managed access, security terms and dual-use limits; organizations apply."
    },
    {
      "id": "openai-gpt-rosalind-6-system-card",
      "title": "GPT-Rosalind-6 System Card Addendum",
      "publisher": "OpenAI",
      "url": "https://deploymentsafety.openai.com/gpt-rosalind-6",
      "sourceType": "official",
      "verifiedAt": "2026-11-05",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "benchmark"
      ],
      "reuseNote": "Published 2026-10-30. GPT-Rosalind-6 is a frontier reasoning model for biology, drug discovery and translational medicine on the same weights as GPT-6 Astra, with fewer refusals on dual-use biology queries, offered only through trusted access after review of each applicant. OpenAI treats it as High capability in the biological and chemical domain and reports MedChemBench v7, GeneBench Pro v13 and LifeSciBench Gold against GPT-Rosalind-5.5."
    },
    {
      "id": "openai-pricing-life-sciences",
      "title": "OpenAI API pricing: life sciences models",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/pricing#life-sciences-models",
      "sourceType": "official",
      "verifiedAt": "2026-11-05",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "gpt-6-rosalind costs 15 USD input, 75 USD output and 1.50 USD cached input per 1M tokens on the Standard tier. Access is limited to approved internal research through the trusted-access program."
    },
    {
      "id": "openai-gpt-6-cyber-release",
      "title": "GPT-6 Cyber: Advancing the frontier of aligned cybersecurity agents",
      "publisher": "OpenAI",
      "url": "https://openai.com/index/introducing-gpt-6-cyber/",
      "sourceType": "official",
      "verifiedAt": "2026-11-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "benchmark"
      ],
      "reuseNote": "Published 2026-11-05. GPT-6 Cyber is available through the API and Codex to a select group of approved Daybreak Red customers first, then to all Daybreak Red customers; GPT-5.6 Cyber becomes available to all of them. OpenAI classifies GPT-6 Cyber as Critical in cybersecurity, up from High for GPT-5.6 Cyber. Its charts: ExploitGym 48.5% (GPT-5.6 Cyber 32.1%, GPT-6 Astra 42.4% and GPT-5.6 Sol 30.3% through Daybreak Blue), SRE-Bench 92.8%, and its own RemoteExploitBench and web pentesting evaluations. OpenAI reports exploits for 948 Linux kernel vulnerabilities and 81 high-severity Firefox vulnerabilities."
    },
    {
      "id": "openai-gpt-6-cyber-system-card",
      "title": "GPT-6 Cyber System Card",
      "publisher": "OpenAI",
      "url": "https://deploymentsafety.openai.com/gpt-6-cyber",
      "sourceType": "official",
      "verifiedAt": "2026-11-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "benchmark"
      ],
      "reuseNote": "Published 2026-11-05. GPT-6 Cyber is a cyber-specialized variant of GPT-6 Astra for authorized security work, with the same Preparedness designations: Critical in cybersecurity, High in biology and chemistry. Access goes through Daybreak Red with identity verification, single sign-on, multi-factor authentication and API IP allowlisting. Results at the highest tested reasoning budget against GPT-6 Astra and GPT-5.6 Sol: ExploitBench, SEC-Bench Pro, SRE-Bench, ExploitGym, plus internal and Irregular evaluations."
    },
    {
      "id": "openai-gpt-6-cyber-model",
      "title": "GPT-6-Cyber model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-6-cyber",
      "sourceType": "official",
      "verifiedAt": "2026-11-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "gpt-6-cyber needs Daybreak Red access and approval. 1,050,000-token context window, 128,000 max output tokens, text and image input, text output. Standard short-context prices per 1M tokens: 20 USD input, 2 USD cached input, 100 USD output; cache writes 25 USD for 30 minutes or 50 USD for 12 hours."
    },
    {
      "id": "openai-gpt-5-6-cyber-model",
      "title": "GPT-5.6 Cyber model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-5.6-cyber",
      "sourceType": "official",
      "verifiedAt": "2026-11-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "gpt-5.6-cyber is an alias for purpose-trained cybersecurity models for approved defenders, available to all Daybreak Red customers. 400,000-token context window, 128,000 max output tokens, knowledge cutoff 2026-02-16. Prices per 1M tokens: 12.50 USD input, 1.25 USD cached input, 75 USD output."
    },
    {
      "id": "openai-daybreak-expansion",
      "title": "Expanding Daybreak as the Cyber Defense Window Narrows",
      "publisher": "OpenAI",
      "url": "https://openai.com/index/expanding-daybreak-as-the-cyber-defense-window-narrows/",
      "sourceType": "official",
      "verifiedAt": "2026-11-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Published 2026-08-10. OpenAI splits Daybreak into Daybreak Blue and Daybreak Red and introduces GPT-5.6-Cyber, built on GPT-5.6 Sol and available through Daybreak Red, rated High but not Critical in cybersecurity. Its internal advanced-request evaluation: 95.0% completed, against 1.5% for GPT-5.6 Sol and 57.3% for GPT-5.5-Cyber."
    },
    {
      "id": "openai-gpt-image-2-6-sunburst-model",
      "title": "GPT-Image-2.6 Sunburst model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-image-2.6-sunburst",
      "sourceType": "official",
      "verifiedAt": "2026-11-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "gpt-image-2.6-sunburst generates and edits images from text and image inputs, for detailed compositions and editing. Per 1M tokens: text input 5 USD (cached 1.25 USD), image input 8 USD (cached 2 USD), image output 30 USD."
    },
    {
      "id": "openai-gpt-image-2-6-flare-model",
      "title": "GPT-Image-2.6 Flare model",
      "publisher": "OpenAI",
      "url": "https://developers.openai.com/api/docs/models/gpt-image-2.6-flare",
      "sourceType": "official",
      "verifiedAt": "2026-11-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "gpt-image-2.6-flare generates and edits images from text and image inputs, for fast iteration. Per 1M tokens: text input 5 USD (cached 1.25 USD), image input 8 USD (cached 2 USD), image output 30 USD."
    },
    {
      "id": "openai-images-2-6",
      "title": "Introducing Images 2.6",
      "publisher": "OpenAI",
      "url": "https://openai.com/index/introducing-images-2-6/",
      "sourceType": "official",
      "verifiedAt": "2026-11-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Published 2026-11-04. Images 2.6 in ChatGPT, Codex and the API generates images 50% faster than before; the API offers GPT-Image-2.6 Flare (as little as six seconds per image at low quality) and GPT-Image-2.6 Sunburst (higher quality and consistency across edits), with animated GIFs, SVG export, image markup and a complexity parameter."
    },
    {
      "id": "google-gemini-changelog-2026-11-05",
      "title": "Gemini API release notes, November 5, 2026",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/changelog#11-05-2026",
      "sourceType": "official",
      "verifiedAt": "2026-11-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity"
      ],
      "reuseNote": "Gemini Omni 1.2 Flash (gemini-omni-1.2-flash) generally available: up to 15-second video generation and conversational editing, character consistency across up to 6 reference images, 30% lower video output pricing and doubled paid-tier rate limits. From 2026-12-03, requests to gemini-omni-1.1-flash route to gemini-omni-1.2-flash at its rates."
    },
    {
      "id": "google-pricing-omni-1-2-flash",
      "title": "Gemini Developer API pricing: Gemini Omni 1.2 Flash",
      "publisher": "Google",
      "url": "https://ai.google.dev/gemini-api/docs/pricing#gemini-omni-1.2-flash",
      "sourceType": "official",
      "verifiedAt": "2026-11-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "pricing"
      ],
      "reuseNote": "Paid tier only. Per 1M tokens: 1.50 USD input (text, image, video), 9 USD text output, 12 USD video output, billed at 5,792 tokens per second of 720p video: about 0.07 USD a second at 720p, 0.023 USD at 360p, 0.104 USD at 1080p and 0.209 USD at 4K with upsampling."
    },
    {
      "id": "thinking-machines-inkling-2-small-release",
      "title": "Introducing Inkling 2 Small",
      "publisher": "Thinking Machines Lab",
      "url": "https://thinkingmachines.ai/news/inkling-2-small/",
      "sourceType": "official",
      "verifiedAt": "2026-11-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "benchmark"
      ],
      "reuseNote": "Published 2026-11-05. Inkling 2 Small is a mixture-of-experts model with 594B total and 17B active parameters, image input and variable effort, available first on Tinker with closed weights; Thinking Machines intends to publish the full weights under Apache 2.0 the following week, after external safety testing."
    },
    {
      "id": "thinking-machines-inkling-2-small-card",
      "title": "Inkling 2 Small model card",
      "publisher": "Thinking Machines Lab",
      "url": "https://thinkingmachines.ai/model-card/inkling-2-small/",
      "sourceType": "official",
      "verifiedAt": "2026-11-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "benchmark"
      ],
      "reuseNote": "Released 2026-11-05 by Thinking Machines Lab, Inc. A 48-layer decoder-only mixture-of-experts transformer (8 of 384 experts per token), 594B total and 17B active parameters, text and image input, text output, a context window of up to 1M tokens, offered through the Tinker platform under the Tinker Terms of Use. Benchmark table against twelve open-weight and closed-weight models."
    },
    {
      "id": "tinker-models-pricing",
      "title": "Tinker models and pricing",
      "publisher": "Thinking Machines Lab",
      "url": "https://tinker-docs.thinkingmachines.ai/tinker/models/models_and_pricing/",
      "sourceType": "official",
      "verifiedAt": "2026-11-06",
      "datasetVersion": null,
      "license": null,
      "topics": [
        "identity",
        "pricing"
      ],
      "reuseNote": "Inkling 2 Small (thinkingmachines/Inkling-2-Small) on Tinker, per 1M tokens: prefill 0.30 USD (0.03 USD cached), sampling 1.00 USD, training 1.50 USD at 64K context or 3.00 USD for the 256K variant."
    },
    {
      "id": "arena-leaderboard-dataset-2026-11-06",
      "title": "Arena leaderboard dataset (lmarena-ai/leaderboard-dataset)",
      "publisher": "Arena",
      "url": "https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset/tree/5a7e6bb84a93f55df56b198be8e56e53755a216a",
      "sourceType": "independent",
      "verifiedAt": "2026-11-06",
      "datasetVersion": "lmarena-ai/leaderboard-dataset@5a7e6bb84a93f55df56b198be8e56e53755a216a",
      "license": "CC-BY-4.0",
      "topics": [
        "benchmark"
      ],
      "reuseNote": "Ratings from blind pairwise human votes, published by Arena (formerly LMArena) in its public leaderboard dataset and pinned here to one commit. The overall category of the latest split is used; for text, the style-controlled leaderboard, which is the default view on arena.ai, and for code, the WebDev leaderboard of Code Arena. A model maps only when its Arena row names the same model; where Arena lists several quality or resolution settings of one model, the best rated row is used and recorded in sourceRow. Ratings are rounded to whole points. Credit: Arena, lmarena-ai/leaderboard-dataset on Hugging Face, CC BY 4.0."
    }
  ],
  "recommendations": [
    {
      "useCase": "agents",
      "modelIds": [
        "gpt-56-sol",
        "claude-sonnet-5",
        "gemini-36-flash"
      ],
      "summary": "Start a current agent-workflow evaluation with GPT-5.6 Sol, Claude Sonnet 5, and Gemini 3.6 Flash.",
      "rationale": [
        "All three active API records are positioned by their providers for agentic or tool-using workflows.",
        "GPT-5.6 Sol and Gemini 3.6 Flash publish context windows above one million tokens, while Claude Sonnet 5 launches with a lower introductory input price than GPT-5.6 Sol."
      ],
      "conditions": [
        "Test the models with the actual tools, permissions, retry rules, and human approval steps used in production."
      ],
      "limitations": [
        "The pinned LiveBench snapshot predates GPT-5.6 Sol and Gemini 3.6 Flash and does not measure tool-use success, latency, uptime, or end-to-end agent reliability; Claude Sonnet 5 pricing changes after 2026-08-31."
      ],
      "sourceIds": [
        "openai-gpt-5-6-release",
        "openai-gpt-5-6-models",
        "anthropic-claude-sonnet-5-release",
        "google-gemini-3-6-model",
        "google-gemini-3-6-latest"
      ]
    },
    {
      "useCase": "coding",
      "modelIds": [
        "gpt-56-sol",
        "claude-fable-5"
      ],
      "summary": "Test GPT-5.6 Sol and Claude Fable 5 as current coding-focused API options before choosing a model for repository work.",
      "rationale": [
        "Both providers position these active records for complex coding or software-engineering workflows.",
        "Their published million-token-class context windows support evaluation on large repositories and long-running work."
      ],
      "conditions": [
        "Evaluate repository navigation, code edits, test execution, and review quality on your own languages and tooling."
      ],
      "limitations": [
        "The catalog does not include a compatible repository-level coding benchmark for either model, and the pinned LiveBench snapshot predates both models."
      ],
      "sourceIds": [
        "openai-gpt-5-6-release",
        "openai-gpt-5-6-models",
        "anthropic-claude-fable-5-release",
        "anthropic-models-overview-current"
      ]
    },
    {
      "useCase": "analysis",
      "modelIds": [
        "gpt-55",
        "claude-fable-5"
      ],
      "summary": "Compare GPT-5.5 and Claude Fable 5 for current high-capability analysis, then test both against your review standard.",
      "rationale": [
        "Both providers position these active API records for demanding reasoning or professional knowledge work.",
        "Both records publish context windows of at least 1,000,000 tokens."
      ],
      "conditions": [
        "Use representative documents, expected citations, and a defined human review rubric for the production trial."
      ],
      "limitations": [
        "The pinned LiveBench snapshot does not provide a compatible Claude Fable 5 result and does not establish factuality, domain expertise, citation quality, latency, or production reliability."
      ],
      "sourceIds": [
        "openai-gpt-5-5-model",
        "anthropic-claude-fable-5-release",
        "anthropic-models-overview-current"
      ]
    },
    {
      "useCase": "multimodal",
      "modelIds": [
        "gpt-56-terra",
        "gemini-36-flash",
        "grok-45"
      ],
      "summary": "Compare current API options from OpenAI, Google, and SpaceXAI when the workload combines text with visual inputs.",
      "rationale": [
        "Gemini 3.6 Flash records text, image, audio, and video inputs; GPT-5.6 Terra and Grok 4.5 record text and image inputs.",
        "The three current records span published context windows from 500,000 to 1,050,000 tokens and different input-price points."
      ],
      "conditions": [
        "Build an evaluation set from the image, audio, or video formats and document layouts the system will receive."
      ],
      "limitations": [
        "The dataset has no compatible multimodal quality benchmark for this shortlist and does not record modality-specific pricing."
      ],
      "sourceIds": [
        "openai-gpt-5-6-models",
        "google-gemini-3-6-model",
        "google-gemini-3-6-latest",
        "xai-grok-4-5-model"
      ]
    },
    {
      "useCase": "budget",
      "modelIds": [
        "gemini-35-flash-lite",
        "gpt-41-nano"
      ],
      "summary": "Compare current Gemini 3.5 Flash-Lite with GPT-4.1 nano for cost-sensitive API trials using published list prices.",
      "rationale": [
        "Gemini 3.5 Flash-Lite records $0.30 input and $2.50 output per million tokens, while GPT-4.1 nano records $0.10 input and $0.40 output.",
        "Both active records accept multimodal input and publish context windows above one million tokens."
      ],
      "conditions": [
        "Estimate full task cost with representative input size, output size, retries, caching, review, and supporting infrastructure."
      ],
      "limitations": [
        "List price does not include every production cost, and this dataset has no compatible quality score for either record."
      ],
      "sourceIds": [
        "google-gemini-3-5-flash-lite-model",
        "google-gemini-3-5-flash-lite-pricing",
        "openai-gpt-4-1-nano-model",
        "openai-pricing"
      ]
    }
  ],
  "reportedTables": [
    {
      "id": "google-gemini-4-carbon-launch",
      "providerId": "google",
      "title": "Gemini 4 Carbon launch table",
      "publishedAt": "2026-10-28",
      "sourceId": "google-gemini-4-carbon-release",
      "modelIds": [
        "gemini-4-carbon",
        "gemini-4-argon",
        "gpt-6-astra",
        "claude-opus-55",
        "claude-fable-5-1"
      ],
      "note": "Google's table from the Gemini 4 Carbon announcement, evaluated without guardrails; Google publishes its methodology at deepmind.google/models/evals-methodology/gemini-4-carbon. The cybersecurity rows, Google's headline, come first. Left out here: the Artificial Analysis and Vals indexes and the two rows Artificial Analysis ran (Terminal-Bench 4.0 and Terminal-Bench-Science 0.1).",
      "rows": [
        {
          "benchmark": "SEC-bench Pro",
          "unit": "percent",
          "scores": [
            88.2,
            83.4,
            85.4,
            null,
            null
          ]
        },
        {
          "benchmark": "ExploitGym",
          "unit": "percent",
          "scores": [
            54.8,
            42,
            42.4,
            34.5,
            null
          ]
        },
        {
          "benchmark": "CWE-bench",
          "unit": "percent",
          "scores": [
            71,
            68,
            68,
            67,
            58
          ]
        },
        {
          "benchmark": "SWE-sweep",
          "unit": "percent",
          "scores": [
            9.6,
            4.3,
            9.5,
            3.6,
            2.9
          ]
        },
        {
          "benchmark": "FrontierSWE 2",
          "unit": "percent",
          "scores": [
            61.2,
            55,
            65.5,
            62.3,
            56.3
          ]
        },
        {
          "benchmark": "Agents' Last Exam v1.1",
          "setting": "score",
          "unit": "percent",
          "scores": [
            77.4,
            69,
            73.1,
            71.6,
            null
          ]
        },
        {
          "benchmark": "OSWorld-2.1",
          "setting": "offline subset",
          "unit": "percent",
          "scores": [
            81.5,
            74.7,
            80.9,
            87.2,
            76.8
          ]
        },
        {
          "benchmark": "AutomationBench",
          "setting": "Zapier",
          "unit": "percent",
          "scores": [
            53.9,
            51.3,
            41.4,
            42.5,
            31.4
          ]
        },
        {
          "benchmark": "Vending-Bench 2",
          "setting": "bank balance",
          "unit": "usd",
          "scores": [
            16484,
            13718,
            15515,
            9235,
            5422
          ]
        },
        {
          "benchmark": "Chartography",
          "unit": "percent",
          "scores": [
            72.5,
            71.6,
            71,
            66.3,
            46.2
          ]
        },
        {
          "benchmark": "Blueprint-Bench 2",
          "setting": "similarity score",
          "unit": "percent",
          "scores": [
            58.6,
            54.4,
            49.7,
            51.2,
            41.9
          ]
        },
        {
          "benchmark": "GraphWalks",
          "setting": "256K to 1M tokens, BFS F1",
          "unit": "percent",
          "scores": [
            96.2,
            84.2,
            71.8,
            66.8,
            65
          ]
        },
        {
          "benchmark": "RiemannBench",
          "unit": "percent",
          "scores": [
            80,
            76,
            72,
            69.6,
            65.6
          ]
        }
      ]
    },
    {
      "id": "moonshot-kimi-k3-1-pro-card",
      "providerId": "kimi",
      "title": "Kimi K3.1 Pro model card",
      "publishedAt": "2026-11-02",
      "sourceId": "moonshot-kimi-k3-1-pro-weights",
      "modelIds": [
        "kimi-k3-1-pro",
        "claude-fable-5-1",
        "gpt-6-astra",
        "claude-opus-55",
        "deepseek-v41-flash",
        "glm-5-3"
      ],
      "note": "Moonshot's table from the Kimi K3.1 Pro model card, every model at its highest reasoning effort; rivals' scores are the highest their labs or the benchmarks' leaderboards report, unless Moonshot ran them (its footnotes say which). DeepSeek-V4.1-Flash is the 1017 build. Left out here: Moonshot's internal benchmarks, the rows Artificial Analysis runs and the hallucination rows, where lower is better.",
      "rows": [
        {
          "benchmark": "Terminal-Bench 4.0",
          "unit": "percent",
          "scores": [
            65.7,
            57.9,
            58.2,
            66.4,
            42.3,
            41.8
          ]
        },
        {
          "benchmark": "Terminal-Bench-Science 0.1",
          "unit": "percent",
          "scores": [
            38.1,
            40,
            68.1,
            63.3,
            15.7,
            8.1
          ]
        },
        {
          "benchmark": "SWE Marathon v1.1",
          "unit": "percent",
          "scores": [
            51.7,
            45.6,
            42.5,
            46.3,
            40,
            42.5
          ]
        },
        {
          "benchmark": "DeepSWE v1.1",
          "unit": "percent",
          "scores": [
            75.5,
            67.4,
            74.1,
            74.2,
            77,
            66.9
          ]
        },
        {
          "benchmark": "ProgramBench",
          "setting": "tasks at least 95% solved",
          "unit": "percent",
          "scores": [
            31.5,
            53.5,
            50,
            65,
            29.1,
            19
          ]
        },
        {
          "benchmark": "PostTrainBench v1.2",
          "unit": "percent",
          "scores": [
            36.5,
            44.6,
            41.9,
            43.8,
            30.5,
            34.5
          ]
        },
        {
          "benchmark": "BenchCAD",
          "setting": "without tools",
          "unit": "percent",
          "scores": [
            72.8,
            60.6,
            null,
            73,
            41.2,
            null
          ]
        },
        {
          "benchmark": "BenchCAD",
          "setting": "with tools",
          "unit": "percent",
          "scores": [
            92.7,
            92.6,
            95.9,
            96.2,
            81.6,
            null
          ]
        },
        {
          "benchmark": "HLE-Full",
          "setting": "with tools",
          "unit": "percent",
          "scores": [
            70,
            65.6,
            57.2,
            67.7,
            58.8,
            62.5
          ]
        },
        {
          "benchmark": "HLE-Full",
          "setting": "without tools",
          "unit": "percent",
          "scores": [
            51.1,
            57.6,
            53.6,
            59.8,
            42.6,
            43
          ]
        },
        {
          "benchmark": "HLE-Diamond",
          "unit": "percent",
          "scores": [
            39.4,
            55.1,
            65.8,
            60.9,
            21.6,
            25.1
          ]
        },
        {
          "benchmark": "GPQA-Diamond",
          "unit": "percent",
          "scores": [
            94.2,
            93.7,
            96.1,
            93.9,
            92.2,
            91.7
          ]
        },
        {
          "benchmark": "BrowseComp",
          "unit": "percent",
          "scores": [
            92.1,
            null,
            91.5,
            90.8,
            null,
            null
          ]
        },
        {
          "benchmark": "BioMysteryBench",
          "unit": "percent",
          "scores": [
            80.6,
            81.6,
            78.9,
            81.9,
            80,
            79.6
          ]
        },
        {
          "benchmark": "APEX-Agents",
          "unit": "percent",
          "scores": [
            69.4,
            68.6,
            64.7,
            73.5,
            69.8,
            56.6
          ]
        },
        {
          "benchmark": "Agents' Last Exam",
          "setting": "all tasks",
          "unit": "percent",
          "scores": [
            43.2,
            null,
            41.8,
            42.6,
            37.7,
            26.2
          ]
        },
        {
          "benchmark": "Agents' Last Exam",
          "setting": "command-line tasks",
          "unit": "percent",
          "scores": [
            44.2,
            null,
            44.2,
            40.7,
            36,
            29.4
          ]
        },
        {
          "benchmark": "OSWorld 2.1",
          "unit": "percent",
          "scores": [
            79.6,
            80.7,
            75.8,
            81.8,
            69.3,
            null
          ]
        },
        {
          "benchmark": "OfficeQA Pro v2",
          "unit": "percent",
          "scores": [
            64.4,
            54.5,
            54.4,
            62.2,
            40.4,
            31.1
          ]
        },
        {
          "benchmark": "Frontier Finance",
          "unit": "percent",
          "scores": [
            58.6,
            60.8,
            45,
            56.7,
            54.5,
            57.5
          ]
        },
        {
          "benchmark": "HANDBOOK.md",
          "unit": "percent",
          "scores": [
            44.6,
            38.8,
            null,
            41.5,
            28.9,
            20.8
          ]
        },
        {
          "benchmark": "Harvey Legal Agent",
          "unit": "percent",
          "scores": [
            23.7,
            19.1,
            null,
            null,
            21,
            null
          ]
        },
        {
          "benchmark": "GDP.pdf",
          "unit": "percent",
          "scores": [
            34.6,
            27.6,
            34.2,
            30.6,
            17.2,
            11.2
          ]
        },
        {
          "benchmark": "GDP.pdf",
          "setting": "with Python",
          "unit": "percent",
          "scores": [
            85.7,
            83.8,
            85.5,
            83.9,
            73,
            null
          ]
        },
        {
          "benchmark": "MMMU-Pro",
          "unit": "percent",
          "scores": [
            84.1,
            86.3,
            86.9,
            87.7,
            80.6,
            null
          ]
        },
        {
          "benchmark": "MMMU-Pro",
          "setting": "with Python",
          "unit": "percent",
          "scores": [
            90,
            88.4,
            87.8,
            89.8,
            82.5,
            null
          ]
        },
        {
          "benchmark": "CharXiv (RQ)",
          "unit": "percent",
          "scores": [
            89,
            88.8,
            90,
            92.1,
            77,
            null
          ]
        },
        {
          "benchmark": "CharXiv (RQ)",
          "setting": "with Python",
          "unit": "percent",
          "scores": [
            93.4,
            92.7,
            92.8,
            93.8,
            85.6,
            null
          ]
        },
        {
          "benchmark": "ChartoGraphy",
          "setting": "with Python",
          "unit": "percent",
          "scores": [
            90.2,
            88.4,
            91,
            89,
            83,
            null
          ]
        },
        {
          "benchmark": "BabyVision",
          "setting": "with Python",
          "unit": "percent",
          "scores": [
            95.4,
            95.1,
            96.6,
            97.7,
            92.5,
            null
          ]
        },
        {
          "benchmark": "ZeroBench main",
          "setting": "pass@5",
          "unit": "percent",
          "scores": [
            39,
            35,
            52,
            43,
            23,
            null
          ]
        },
        {
          "benchmark": "ZeroBench main",
          "setting": "with Python, pass@5",
          "unit": "percent",
          "scores": [
            63,
            59,
            58,
            61,
            51,
            null
          ]
        },
        {
          "benchmark": "PerceptionBench",
          "unit": "percent",
          "scores": [
            68.3,
            63,
            77.6,
            72.1,
            47.7,
            null
          ]
        },
        {
          "benchmark": "ExploitBench",
          "unit": "percent",
          "scores": [
            74.7,
            83,
            100,
            91,
            null,
            54.4
          ]
        },
        {
          "benchmark": "ExploitGym",
          "unit": "percent",
          "scores": [
            20.7,
            30.4,
            42.4,
            34.5,
            null,
            15
          ]
        }
      ]
    },
    {
      "id": "tencent-hy4-launch",
      "providerId": "tencent",
      "title": "Hy4 launch table",
      "publishedAt": "2026-11-04",
      "sourceId": "tencent-hy4-release",
      "modelIds": [
        "hy4",
        "glm-5-3",
        "deepseek-v41-flash",
        "kimi-k3-1-pro",
        "gpt-61-sol",
        "claude-opus-55"
      ],
      "note": "Tencent's appendix to the Hy4 announcement, every model at its highest reasoning setting. Where Tencent lists a lab's reported score beside its own run, the score Tencent measured is shown. DeepSeek-V4.1-Flash is the 1017 build. Left out here: Tencent's internal benchmarks, the Hy4 preview column and GLM-5.3 on HLE, which Tencent ran on the text-only subset.",
      "rows": [
        {
          "benchmark": "DeepSWE v1.1",
          "unit": "percent",
          "scores": [
            74.9,
            68.1,
            73.8,
            76.3,
            72.7,
            72.6
          ]
        },
        {
          "benchmark": "SWE-Marathon v1.1",
          "unit": "percent",
          "scores": [
            43.8,
            35.6,
            46.3,
            53.8,
            39.4,
            44.4
          ]
        },
        {
          "benchmark": "ProgramBench",
          "setting": "almost solved",
          "unit": "percent",
          "scores": [
            30,
            18,
            29,
            24.5,
            46,
            56.5
          ]
        },
        {
          "benchmark": "Terminal-Bench 4.0",
          "unit": "percent",
          "scores": [
            44.4,
            42.9,
            40,
            66.8,
            55.5,
            66.4
          ]
        },
        {
          "benchmark": "FrontierCode 2.0",
          "unit": "percent",
          "scores": [
            31,
            21.6,
            28.5,
            25.5,
            40.6,
            46.4
          ]
        },
        {
          "benchmark": "APEX-Agents v1.1",
          "setting": "pass@1",
          "unit": "percent",
          "scores": [
            75.2,
            64.3,
            72.9,
            68.2,
            61.7,
            74.8
          ]
        },
        {
          "benchmark": "Agents' Last Exam v1.1",
          "setting": "pass rate",
          "unit": "percent",
          "scores": [
            41,
            26.2,
            35.3,
            43.4,
            40.2,
            43.4
          ]
        },
        {
          "benchmark": "AutomationBench v1.0.6",
          "unit": "percent",
          "scores": [
            56.2,
            54.1,
            54.3,
            65.6,
            51.4,
            58.8
          ]
        },
        {
          "benchmark": "HLE-Full",
          "setting": "with tools",
          "unit": "percent",
          "scores": [
            59.1,
            null,
            57.1,
            61.8,
            60.6,
            66.2
          ]
        },
        {
          "benchmark": "Terminal-Bench-Science 0.1",
          "unit": "percent",
          "scores": [
            19.8,
            10,
            13.7,
            41.7,
            60.9,
            56.2
          ]
        },
        {
          "benchmark": "PostTrainBench v1.2",
          "unit": "percent",
          "scores": [
            35.2,
            31.6,
            29.4,
            36.2,
            38.2,
            44.2
          ]
        },
        {
          "benchmark": "OSWorld 2.1",
          "setting": "full, partial score",
          "unit": "percent",
          "scores": [
            78.9,
            null,
            68.4,
            79.6,
            75.2,
            81.8
          ]
        },
        {
          "benchmark": "MMMU-Pro",
          "unit": "percent",
          "scores": [
            78.6,
            null,
            76.7,
            83.4,
            84.9,
            88.2
          ]
        },
        {
          "benchmark": "ZeroBench main",
          "setting": "with tools, pass@5",
          "unit": "percent",
          "scores": [
            55.4,
            null,
            51,
            56.9,
            65.3,
            63.5
          ]
        }
      ]
    },
    {
      "id": "openai-gpt-rosalind-6-card",
      "providerId": "openai",
      "title": "GPT-Rosalind-6 system card addendum",
      "publishedAt": "2026-10-30",
      "sourceId": "openai-gpt-rosalind-6-system-card",
      "modelIds": [
        "gpt-rosalind-6"
      ],
      "note": "OpenAI's life sciences evaluations at the highest reasoning budget it tested. OpenAI compares them only with GPT-Rosalind-5.5, which is not in this catalog: 28.9%, 16.7% and 44.0% on the same problem sets.",
      "rows": [
        {
          "benchmark": "MedChemBench v7",
          "setting": "mean rubric score, 99 problems",
          "unit": "percent",
          "scores": [
            49.6
          ]
        },
        {
          "benchmark": "GeneBench Pro v13",
          "setting": "single-attempt success, 129 problems",
          "unit": "percent",
          "scores": [
            37.7
          ]
        },
        {
          "benchmark": "LifeSciBench Gold",
          "setting": "average score, 94 problems",
          "unit": "percent",
          "scores": [
            61.3
          ]
        }
      ]
    },
    {
      "id": "openai-gpt-6-cyber-card",
      "providerId": "openai",
      "title": "GPT-6 Cyber system card",
      "publishedAt": "2026-11-05",
      "sourceId": "openai-gpt-6-cyber-system-card",
      "modelIds": [
        "gpt-6-cyber",
        "gpt-6-astra",
        "gpt-56-sol"
      ],
      "note": "OpenAI's system card results at the highest reasoning budget it tested. The GPT-6 Astra and GPT-5.6 Sol scores match the runs OpenAI's launch post labels Daybreak Blue, the access tier without their cyber guardrails. OpenAI warns that ExploitBench's historical vulnerabilities may have been seen in training. ExploitGym, where the system card gives the same scores, is in the launch post table. Left out here: evaluations OpenAI built itself (ExploitBench Internal Port, ExploitGym Hardened, ExpertSWE) and Irregular's private suites.",
      "rows": [
        {
          "benchmark": "ExploitBench",
          "setting": "mean capability score, normalized to arbitrary code execution",
          "unit": "percent",
          "scores": [
            100,
            100,
            78.5
          ]
        },
        {
          "benchmark": "SEC-Bench Pro",
          "setting": "pass@1",
          "unit": "percent",
          "scores": [
            87,
            85.4,
            79.1
          ]
        },
        {
          "benchmark": "SRE-Bench",
          "setting": "all six objectives, pass@1",
          "unit": "percent",
          "scores": [
            92.8,
            88,
            55.9
          ]
        }
      ]
    },
    {
      "id": "openai-gpt-6-cyber-launch",
      "providerId": "openai",
      "title": "GPT-6 Cyber launch post",
      "publishedAt": "2026-11-05",
      "sourceId": "openai-gpt-6-cyber-release",
      "modelIds": [
        "gpt-6-cyber",
        "gpt-56-cyber",
        "gpt-6-astra",
        "gpt-56-sol"
      ],
      "note": "OpenAI's chart from the GPT-6 Cyber announcement. GPT-6 Astra and GPT-5.6 Sol ran with Daybreak Blue access; with its standard safeguards GPT-6 Astra refuses these tasks and scores 0%. Left out here: RemoteExploitBench and the web pentesting evaluation, which OpenAI built itself, the honeypot evaluation, where lower is better, and SRE-Bench, which is in the system card table.",
      "rows": [
        {
          "benchmark": "ExploitGym",
          "setting": "strict supplied-vulnerability pass@1",
          "unit": "percent",
          "scores": [
            48.5,
            32.1,
            42.4,
            30.3
          ]
        }
      ]
    },
    {
      "id": "thinking-machines-inkling-2-small-card",
      "providerId": "thinking-machines",
      "title": "Inkling 2 Small model card",
      "publishedAt": "2026-11-05",
      "sourceId": "thinking-machines-inkling-2-small-card",
      "modelIds": [
        "inkling-2-small",
        "glm-5-3-flash",
        "mimo-v2-6-flash",
        "minimax-m3-1-flash",
        "deepseek-v41-flash",
        "muse-spark-open",
        "glm-5-3",
        "mimo-v2-6-pro",
        "mistral-large-4",
        "kimi-k3",
        "gpt-6-luna",
        "gemini-38-flash",
        "claude-haiku-55"
      ],
      "note": "Thinking Machines' table from the Inkling 2 Small model card, run at effort 0.99 (xhigh) with a 1M-token trajectory limit; where a benchmark's owner publishes scores (Datacurve, Scale AI, Surge, Mercor, Thoughtful, Vals), Thinking Machines uses them. DeepSeek-V4.1-Flash is the 1017 build. Left out here: the Artificial Analysis and Vals indexes, the rows Artificial Analysis ran (FrontierCode, Terminal-Bench, GDP.pdf, MMMU-Pro, HLE, CritPt, AA-LCR, AA-Omniscience, AutomationBench, EnterpriseOps-Gym, Tau 3 Banking), Arena and DesignArena ratings, PMPP-Hard (not a percentage), the safety rows, and JobBench, DAYJOB and Humanity's Sixth Sense, which we could not confirm are public.",
      "rows": [
        {
          "benchmark": "DeepSWE v1.1",
          "unit": "percent",
          "scores": [
            72.1,
            63.4,
            67.9,
            73.7,
            77,
            60,
            69,
            71.9,
            61.7,
            68.5,
            66.6,
            73.8,
            59.9
          ]
        },
        {
          "benchmark": "Vibe Code Bench v1.1",
          "unit": "percent",
          "scores": [
            81.6,
            30.8,
            79,
            73.9,
            87.7,
            76.2,
            78.1,
            85.2,
            81.6,
            85,
            81.7,
            78.7,
            90.4
          ]
        },
        {
          "benchmark": "Sudo L7",
          "unit": "percent",
          "scores": [
            14.1,
            null,
            null,
            null,
            null,
            null,
            15,
            null,
            13.9,
            13.9,
            null,
            10.6,
            null
          ]
        },
        {
          "benchmark": "SWE-Marathon v1.1",
          "unit": "percent",
          "scores": [
            38.8,
            38.8,
            null,
            null,
            40,
            null,
            42.5,
            null,
            null,
            48.1,
            28.1,
            null,
            50
          ]
        },
        {
          "benchmark": "SWE-bench Multilingual",
          "unit": "percent",
          "scores": [
            81.2,
            88.9,
            null,
            79.9,
            82.9,
            77.5,
            87.6,
            null,
            79.7,
            91.4,
            78.2,
            85.8,
            83.7
          ]
        },
        {
          "benchmark": "SWE-bench Multimodal v2",
          "unit": "percent",
          "scores": [
            43.1,
            null,
            null,
            null,
            null,
            null,
            null,
            41.3,
            null,
            44.2,
            37.5,
            40.8,
            44
          ]
        },
        {
          "benchmark": "SWE-bench Pro v2",
          "setting": "hard set",
          "unit": "percent",
          "scores": [
            86.3,
            null,
            null,
            84.3,
            null,
            82.4,
            84.3,
            null,
            null,
            88.2,
            null,
            58.8,
            52.9
          ]
        },
        {
          "benchmark": "SWE-bench Pro v2",
          "setting": "full set",
          "unit": "percent",
          "scores": [
            97.7,
            null,
            null,
            96.6,
            null,
            95,
            95.6,
            null,
            null,
            97.7,
            null,
            94.9,
            87.2
          ]
        },
        {
          "benchmark": "SWE-bench Verified",
          "unit": "percent",
          "scores": [
            82.2,
            92,
            null,
            81.1,
            84.6,
            80,
            95.4,
            null,
            80,
            93.4,
            76.1,
            80,
            88.9
          ]
        },
        {
          "benchmark": "PostTrainBench v1.2",
          "unit": "percent",
          "scores": [
            31.9,
            36.8,
            null,
            null,
            30.5,
            null,
            34.5,
            null,
            null,
            31.5,
            null,
            null,
            null
          ]
        },
        {
          "benchmark": "BioMysteryBench",
          "unit": "percent",
          "scores": [
            70,
            null,
            69.3,
            null,
            null,
            null,
            null,
            null,
            null,
            72.6,
            61.5,
            62.2,
            70.4
          ]
        },
        {
          "benchmark": "BrowseComp",
          "setting": "with context management",
          "unit": "percent",
          "scores": [
            90.2,
            null,
            null,
            null,
            null,
            null,
            null,
            null,
            84.5,
            null,
            null,
            null,
            null
          ]
        },
        {
          "benchmark": "DeepSearchQA",
          "unit": "percent",
          "scores": [
            90.9,
            null,
            null,
            null,
            null,
            null,
            null,
            null,
            null,
            95,
            null,
            null,
            null
          ]
        },
        {
          "benchmark": "MCP Atlas",
          "setting": "full set",
          "unit": "percent",
          "scores": [
            85.4,
            null,
            null,
            null,
            null,
            89.2,
            84.2,
            null,
            null,
            82.3,
            null,
            null,
            null
          ]
        },
        {
          "benchmark": "MCP Atlas",
          "setting": "public set",
          "unit": "percent",
          "scores": [
            85,
            82,
            62.9,
            80.4,
            83.2,
            87.5,
            83.6,
            84.4,
            84.5,
            84.2,
            77.8,
            87.5,
            81
          ]
        },
        {
          "benchmark": "EnterpriseBench-CoreCraft",
          "unit": "percent",
          "scores": [
            71.8,
            46.7,
            null,
            null,
            null,
            null,
            51.8,
            null,
            null,
            43.1,
            null,
            58.5,
            null
          ]
        },
        {
          "benchmark": "Finance Agent v2.1",
          "unit": "percent",
          "scores": [
            67.5,
            60.6,
            63.3,
            58.8,
            62.9,
            60.7,
            65,
            64.3,
            61.7,
            60,
            56.8,
            65.8,
            65.4
          ]
        },
        {
          "benchmark": "Excel Modeling Benchmark v1.1",
          "unit": "percent",
          "scores": [
            69.5,
            66.3,
            68.9,
            67.8,
            69.2,
            62,
            69.1,
            67.4,
            68.7,
            71.3,
            70,
            72.9,
            65.5
          ]
        },
        {
          "benchmark": "Tax Agent Bench",
          "unit": "percent",
          "scores": [
            65.8,
            62,
            59.9,
            62.7,
            74.9,
            52.1,
            73.1,
            64.9,
            67.5,
            68.7,
            58.9,
            66.8,
            62.2
          ]
        },
        {
          "benchmark": "APEX-Accounting v1.1",
          "setting": "pass@1",
          "unit": "percent",
          "scores": [
            14.5,
            12.3,
            null,
            13.5,
            12.5,
            13.8,
            13.9,
            null,
            12,
            9.5,
            10.8,
            15.5,
            15.3
          ]
        },
        {
          "benchmark": "APEX-Accounting v1.1",
          "setting": "mean score",
          "unit": "percent",
          "scores": [
            62,
            53.7,
            null,
            60,
            59.3,
            58,
            59.5,
            null,
            56.5,
            56.7,
            53.9,
            58.8,
            62.2
          ]
        },
        {
          "benchmark": "Diligence Bench",
          "setting": "best harness",
          "unit": "percent",
          "scores": [
            58.1,
            null,
            null,
            null,
            50.4,
            59,
            49.4,
            null,
            null,
            null,
            42.1,
            51.1,
            40.8
          ]
        },
        {
          "benchmark": "Harvey LAB v1.1",
          "unit": "percent",
          "scores": [
            21.7,
            10.4,
            13.3,
            7.5,
            14.2,
            25,
            14.6,
            13.8,
            15.8,
            12.9,
            4.2,
            12.1,
            4.2
          ]
        },
        {
          "benchmark": "Legal Research Bench",
          "unit": "percent",
          "scores": [
            46.6,
            45.2,
            38,
            34.1,
            45.2,
            38.5,
            49,
            47.1,
            35.6,
            46.2,
            30.3,
            38.9,
            43.8
          ]
        },
        {
          "benchmark": "PerceptionBench",
          "unit": "percent",
          "scores": [
            64.7,
            null,
            null,
            null,
            47.7,
            null,
            null,
            null,
            null,
            58.5,
            56.2,
            null,
            null
          ]
        },
        {
          "benchmark": "ScreenSpot-Pro",
          "unit": "percent",
          "scores": [
            83.8,
            null,
            null,
            null,
            null,
            null,
            null,
            56.6,
            42.3,
            73.8,
            81.1,
            null,
            null
          ]
        },
        {
          "benchmark": "ScreenSpot-Pro",
          "setting": "with Python",
          "unit": "percent",
          "scores": [
            91.5,
            null,
            null,
            null,
            null,
            null,
            null,
            null,
            null,
            83.2,
            87.6,
            null,
            null
          ]
        },
        {
          "benchmark": "BabyVision",
          "unit": "percent",
          "scores": [
            68.6,
            53.4,
            null,
            null,
            null,
            44.3,
            null,
            null,
            42.3,
            53.4,
            null,
            null,
            null
          ]
        },
        {
          "benchmark": "BabyVision",
          "setting": "with Python",
          "unit": "percent",
          "scores": [
            93.7,
            82.6,
            null,
            91,
            92.5,
            80.2,
            null,
            null,
            71.5,
            85.7,
            null,
            null,
            null
          ]
        },
        {
          "benchmark": "CharXiv (RQ)",
          "unit": "percent",
          "scores": [
            85.6,
            80.2,
            66.2,
            82.5,
            77,
            85.4,
            null,
            74.2,
            83.2,
            84.8,
            81.8,
            86.2,
            84.4
          ]
        },
        {
          "benchmark": "CharXiv (RQ)",
          "setting": "with Python",
          "unit": "percent",
          "scores": [
            87.7,
            89.4,
            null,
            null,
            85.6,
            88.7,
            null,
            null,
            87.9,
            91.3,
            85.3,
            92.1,
            null
          ]
        },
        {
          "benchmark": "Chartography",
          "unit": "percent",
          "scores": [
            36.7,
            16.3,
            7.5,
            12.6,
            17.9,
            35.7,
            null,
            10.8,
            21.2,
            26.6,
            32,
            40.9,
            46.4
          ]
        },
        {
          "benchmark": "Chartography",
          "setting": "with Python",
          "unit": "percent",
          "scores": [
            80.4,
            78,
            null,
            null,
            83,
            63.2,
            null,
            null,
            56.6,
            68.1,
            null,
            82,
            null
          ]
        },
        {
          "benchmark": "MathVision",
          "unit": "percent",
          "scores": [
            92.8,
            82.9,
            null,
            null,
            null,
            87.7,
            null,
            null,
            88.9,
            93.1,
            null,
            null,
            null
          ]
        },
        {
          "benchmark": "SAGE v1.1",
          "unit": "percent",
          "scores": [
            88.6,
            89.3,
            83.3,
            85.3,
            null,
            85.8,
            null,
            84.4,
            null,
            86.8,
            83.4,
            88.2,
            89.8
          ]
        },
        {
          "benchmark": "GPQA-Diamond",
          "unit": "percent",
          "scores": [
            93,
            91.2,
            null,
            86.7,
            92.2,
            89,
            91.7,
            89,
            89.7,
            93.5,
            null,
            95.3,
            89.8
          ]
        },
        {
          "benchmark": "Apex 2025",
          "unit": "percent",
          "scores": [
            63.5,
            null,
            null,
            null,
            null,
            32.3,
            null,
            null,
            null,
            65.6,
            null,
            null,
            null
          ]
        },
        {
          "benchmark": "HMMT Feb 2026",
          "unit": "percent",
          "scores": [
            95.8,
            null,
            null,
            null,
            null,
            93.9,
            null,
            null,
            null,
            97,
            null,
            null,
            null
          ]
        },
        {
          "benchmark": "LongBench v2",
          "unit": "percent",
          "scores": [
            70.6,
            68.2,
            60.7,
            60.9,
            69,
            68.7,
            68.4,
            68,
            65.7,
            69,
            65,
            73.7,
            65.3
          ]
        },
        {
          "benchmark": "ComplexConstraints v1.0",
          "unit": "percent",
          "scores": [
            44.5,
            27.6,
            22.1,
            23.2,
            50.4,
            41.3,
            30.5,
            32.3,
            36.5,
            37.9,
            27.3,
            48.4,
            36.8
          ]
        },
        {
          "benchmark": "CyberBench v1.1",
          "unit": "percent",
          "scores": [
            72.7,
            67.9,
            75.4,
            null,
            null,
            null,
            72.1,
            72.9,
            null,
            75.2,
            76.3,
            43.8,
            75.5
          ]
        },
        {
          "benchmark": "PatchEval",
          "unit": "percent",
          "scores": [
            79.6,
            74.8,
            null,
            null,
            80.4,
            77.8,
            69.6,
            null,
            76.1,
            null,
            null,
            null,
            null
          ]
        }
      ]
    }
  ]
}
