Current model · Meta
Muse Ember benchmarks, pricing and specs
Muse Ember is a Meta model released on November 9, 2026 that takes text, image, audio and video input and produces text. It costs $0.10 per 1M input tokens and $0.50 per 1M output tokens and accepts up to 1,048,576 tokens of context. It scores 1608 Elo on Arena WebDev, #26 of 48 current models (Arena, measured November 9, 2026).
- Per 1M input tokens
- $0.10Listed today
- Per 1M output tokens
- $0.50Listed today
- Context window
- 1.05M#12 of 70 current models
- Arena WebDev
- 1608 Elo#26 of 48 current models
Data verified November 10, 2026 · Dataset 2026-11-10
Specs
Key facts
- Provider
- Meta
- Model family
- Muse Ember
- Release date
- November 9, 2026
- Status
- Current
- Access
- API only
- API model id
- muse-ember-1.0
- Context window
- 1,048,576 tokens
- Max output
- 524,288 tokens
- Input
- text, image, audio, video
- Output
- text
- Task families
- text, multimodal
- Declared capabilities
- reasoning, coding, agentic coding, image understanding, video understanding, audio understanding, document understanding
Charts
Against every current text model
Where Muse Ember lands among the current models of its task, on every independent benchmark, on price and on context.
Where it stands
On Arena WebDev, Muse Ember ranks #26 of 48 current text models. 0 of 61 list a lower price per 1M input tokens. Each bar is the share of current models it beats.
Independent benchmarks
Arena WebDev
#26of 48
1,608 Elo · ahead of 47%
Price and context
Price per 1M input tokens
#1of 61
$0.10 · cheaper than 95%
Price per 1M output tokens
#3of 61
$0.50 · cheaper than 92%
Context window
#12of 70
1.05M tokens · longer than 67%
Price
Muse Ember pricing
It is tied for the cheapest of 61 current text models priced per 1M input tokens.
Benchmarks
Independent benchmark results
Scores come from independent evaluators that ran the model themselves, pinned to a dated snapshot. Rank is among current models in the same snapshot. Meta's own launch figures are in the next section, because they are not measured the same way.
| Benchmark | Score | Rank | Measured | Source |
|---|---|---|---|---|
| Arena WebDev | 1608 Elo | #26 of 48 | November 9, 2026 | Arena · 2026-11-10 |
Provider results
Reported by Meta
Meta's own results for Muse Ember, beside the rivals it chose, as its table lists them. Labs pick the benchmarks and run them their own way, so compare within one table only. Internal benchmarks and other evaluators' indexes are left out.
Muse Ember launch charts, November 9, 2026Meta · 2026-11-10
| Benchmark | Muse Ember | Claude Haiku 5.5 | GPT-6 Luna |
|---|---|---|---|
| Agents' Last ExamLinux subset | 61.6% | 57.0% | 57.1% |
| Terminal-Bench 4.0best published harness | 40.9% | 39.2% | 16.4% |
| OSWorld 2.1partial credit | 69.5% | 68.9% | 50.5% |
| OSWorld 2.1binary | 33.8% | 27.8% | 17.0% |
| PerceptionBenchwith tools | 80.6% | 78.6% | 62.8% |
Meta's charts from the Muse Ember announcement, every model at max reasoning effort and run by Meta. Terminal-Bench 4.0 uses each model's best published harness: mini-swe-agent for Muse Ember, Claude Code for Claude Haiku 5.5 and Codex for GPT-6 Luna. Left out here: the rows Artificial Analysis ran (AutomationBench-AA, Coding Agent Index-AA, FrontierCode 2.0-AA, GDP.pdf-AA and its Intelligence Index) and Meta's internal Personal Superintelligence Index.
Fit
Strengths, cautions and best uses
- Strengths
- Meta's smallest and cheapest Muse model, built for the work agents repeat: tool calls, coding subtasks, routing, extraction and long documents, with text, image, video, audio and PDF input
- $0.10 input, $0.50 output and $0.004 cached input per 1M tokens for requests up to 200K tokens, with a 1,048,576-token context and up to 524,288 output tokens
- In Meta's table at max effort it beats Claude Haiku 5.5 and GPT-6 Luna on Agents' Last Exam (61.6%), Terminal-Bench 4.0 (40.9%), OSWorld 2.1 (69.5%) and PerceptionBench (80.6%)
- Arena WebDev rates it 1608, above Claude Haiku 5.5 (1595) and GPT-6 Luna (1578) (November 9, 2026)
- Cautions
- A request over 200K input tokens is billed in full at three times the rate: $0.30 input, $1.50 output and $0.012 cached input per 1M tokens
- Claude Haiku 5.5 scores higher on FrontierCode 2.0 as run by Artificial Analysis (34.9% against 27.2%), a row in Meta's own launch post
- Arena Text has not rated it yet, and its WebDev rating rests on 1,345 votes so far
- Best for
- Sub-agent work under a larger planner such as Muse Spark or Muse Wonder
- High-volume extraction, classification and routing over images, documents and records
From benchmark to production
Building on Muse Ember?
Public scores rarely hold on your own data. We test it against the alternatives on your real tasks, then build the agent or pipeline around the model that wins.
AI Model Cost AuditLaunch price $100, regular price $150What the audit includes
Alternatives
Alternatives to Muse Ember
The closest current text models by price. The cheapest model that scores as well or better is listed first; failing that, the cheapest one within 3% of the benchmark's scale.
On score and price
Muse Ember beside the 4 alternatives listed below, on Arena WebDev and on blended price per 1M tokens. One of them costs less and scores as well or better.
Arena WebDev
Measured November 9, 2026, higher is better
- Muse Ember
- 1,608
- Claude Haiku 5.5
- 1,595
- GPT-6 Luna
- 1,578
- MiMo-V2.6-Flash
- 1,637
- GPT-4.1 nano
- Not scored
Blended price per 1M tokens
Listed today, lower is better
- Muse Ember
- $0.2
- Claude Haiku 5.5
- $0.2
- GPT-6 Luna
- $0.2
- MiMo-V2.6-Flash
- $0.175
- GPT-4.1 nano
- $0.175
Claude Haiku 5.5
- Same price per 1M input tokens, $0.10
- 13 points lower on Arena WebDev, 1595 vs 1608
- From Anthropic
GPT-6 Luna
- Same price per 1M input tokens, $0.10
- 30 points lower on Arena WebDev, 1578 vs 1608
- From OpenAI
MiMo-V2.6-Flash
- 40% more per 1M input tokens, $0.14 vs $0.10
- 29 points higher on Arena WebDev, 1637 vs 1608
- From Xiaomi MiMo
GPT-4.1 nano
- Same price per 1M input tokens, $0.10
- From OpenAI
Further reading
More on Muse Ember
FAQ
Muse Ember FAQ
How much does Muse Ember cost?
Muse Ember costs $0.10 per 1M input tokens and $0.50 per 1M output tokens at the standard tier. It is tied for the cheapest of 61 current text models priced per 1M input tokens. Source: Meta, verified November 10, 2026.
What is the context window of Muse Ember?
Muse Ember accepts up to 1,048,576 tokens of context. It can return up to 524,288 tokens in one response. That ranks #12 of 70 current text models by context length.
How does Muse Ember score on independent benchmarks?
Muse Ember scores 1608 Elo on Arena WebDev, #26 of 48 current models (Arena, measured November 9, 2026). These are independent evaluations, not Meta's own launch figures.
What benchmark results has Meta published for Muse Ember?
In the Muse Ember launch charts (November 9, 2026), Meta reports 61.6% on Agents' Last Exam (Linux subset), 40.9% on Terminal-Bench 4.0 (best published harness), 69.5% on OSWorld 2.1 (partial credit) and 33.8% on OSWorld 2.1 (binary). These are Meta's own figures, beside the rivals it chose, not independent measurements.
Is Muse Ember open source or API only?
Muse Ember is available through Meta's API only; its weights are not published.
What are the best alternatives to Muse Ember?
The closest current alternatives by price and measured performance are Claude Haiku 5.5 (same price per 1M input tokens, $0.10), GPT-6 Luna (same price per 1M input tokens, $0.10), MiMo-V2.6-Flash (40% more per 1M input tokens, $0.14 vs $0.10) and GPT-4.1 nano (same price per 1M input tokens, $0.10).
Sources
Sources
- Introducing Our Smallest, Most Efficient Model: Muse Ember · Meta · official · verified November 10, 2026
- Muse Ember model page · Meta · official · verified November 10, 2026
- Meta Model API pricing: Muse Ember · Meta · official · verified November 10, 2026
- Arena leaderboard dataset (lmarena-ai/leaderboard-dataset) · Arena · independent · verified November 10, 2026 · CC-BY-4.0
All 262 sources behind the benchmarks, with their versions, licenses and notes, are in the source catalog.
Cite this page
Van Data Team. "Muse Ember Benchmarks, Pricing and Specs." AI Model Benchmark Dataset 2026-11-10. https://vandatateam.com/ai-model-benchmarks/muse-ember
Reuse the data: JSON · CSV · or explore every model in the AI model benchmark comparison.
Production evaluation
Deciding whether Muse Ember fits your workload?
Public benchmarks narrow the shortlist. We test the finalists on your own tasks, data and cost limits, then build the routing and review controls to run the winner in production.
- Evaluation cases and acceptance thresholds tied to real work
- Model routing, fallbacks and cost controls