Skip to main content

Meta sources

13 official pages from Meta support 9 model records in the AI model benchmarks. Official sources come from the model providers themselves: they support each model record as a whole, and no benchmark score comes from them.

Each record names the date it was last verified, its dataset version and license where Meta states one, what it covers and how the data uses it. Missing metadata is not filled or inferred. Every publisher is listed in the source catalog.

Meta 13

  • Introducing Muse Spark Open

    Covers identity · Verified 2026-10-23 · Not versioned · License not stated

    Official announcement, 2026-10-15: 816B total and 42B active parameters, 1M-token context, weights under the Muse Model License.

  • Muse Spark Open model card

    Covers identity · Verified 2026-10-23 · Not versioned · License not stated

    Official open-weights repository: text, image, video and audio in, text out; no first-party API.

  • Introducing Our Smallest, Most Efficient Model: Muse Ember

    Covers identity, pricing, benchmark · Verified 2026-11-10 · Not versioned · License not stated

    Published 2026-11-09. Muse Ember (muse-ember-1.0), the first model trained using Muse Wonder Full, is available on Meta Model API and Muse Code at $0.10 per million input, $0.004 per million cached input and $0.50 per million output tokens up to 200K input tokens, with a 1M-token context and up to 512K output tokens. Its charts compare it at max reasoning with Claude Haiku 5.5 and GPT-6 Luna; rows marked AA were run by Artificial Analysis, the rest by Meta.

  • Muse Ember model page

    Covers identity · Verified 2026-11-10 · Not versioned · License not stated

    Model id muse-ember-1.0, Standard mode only: text, image, video, audio and PDF input, text output, a 1,048,576-token context window and 524,288 output tokens, reasoning effort minimal to max (max by default), on the Responses, Chat Completions and Messages endpoints.

  • Meta Model API pricing: Muse Ember

    Covers pricing · Verified 2026-11-10 · Not versioned · License not stated

    Read 2026-11-10. Muse Ember has two tiers set by the input tokens of a request, cached tokens included: up to 200K input tokens, $0.10 input, $0.004 cached input and $0.50 output per million tokens; above 200K the whole request is billed at $0.30, $0.012 and $1.50. Web search grounding costs $2.50 per 1,000 queries.

The 5 records below: First-party provider documentation used only for the stated model identity, capabilities, lifecycle, access, and pricing.

The 3 records below: Official provider documentation used to verify model identity and release details.

  • Introducing Meta Llama 3

    Covers identity · Verified 2026-08-06 · Not versioned · License not stated

  • Introducing Llama 3.1

    Covers identity · Verified 2026-08-06 · Not versioned · License not stated

  • Llama 3.2

    Covers identity · Verified 2026-08-06 · Not versioned · License not stated

Model pages these records support 9