Muse Ember is Meta's AI model, released 09/11/2026. Meta calls Muse Ember its smallest and most efficient model, built for agent subtasks such as tool calls, coding, routing, extraction and long-document parsing. At max reasoning it beats Claude Haiku 5.5 and GPT-6 Luna on most printed rows but trails Haiku 5.5 on FrontierCode 2.0-AA and ties it on the Artificial Analysis Index. Meta also says it matches Haiku 5.5 on intelligence at about a third of the cost per task.
Visit Muse EmberAI-Powered
Leverages advanced AI technology to deliver cutting-edge capabilities and results.
Fast & Efficient
Optimized performance ensures quick results without compromising on quality.
Purpose-Built
Specifically designed for llms tasks and workflows.
How does Meta Muse Ember score on independent benchmarks?
Meta Muse Ember ranks #15 of 28 models on the Artificial Analysis Intelligence Index (v5.0.1) with a score of 36, as measured on 22 November 2026.
Where does Meta Muse Ember rank against other models?
How good is Meta Muse Ember at agentic coding?
How does Meta Muse Ember do on individual tests?
| Evaluation (run by Artificial Analysis) | Meta Muse Ember | Best of any model |
|---|---|---|
| Terminal-Bench 4.0 | 40.9% | 67.2% |
| FrontierCode 2.0 | 27.2% | 48.8% |
| Humanity's Last Exam | n/a | 74.3% |
| GDPval-AA (Elo) | n/a | 1866 |
How has Meta Muse Ember's score changed over time?
We have tracked Meta Muse Ember on Intelligence Index v5.0.1 since 22/11/2026. A point is added each day the score, coding index or price changes, so this chart fills in over time.
What does the Meta Muse Ember system card report?
Meta calls Muse Ember its smallest and most efficient model, built for agent subtasks such as tool calls, coding, routing, extraction and long-document parsing. At max reasoning it beats Claude Haiku 5.5 and GPT-6 Luna on most printed rows but trails Haiku 5.5 on FrontierCode 2.0-AA and ties it on the Artificial Analysis Index. Meta also says it matches Haiku 5.5 on intelligence at about a third of the cost per task.
These are Meta's own published results (self-reported, with the vendor's harness and settings), so they are not directly comparable with other labs' numbers. Sources: Muse Ember Evaluation Report (methodology) (undefined/11/2026), Muse Ember research blog (09/11/2026). Last checked 22/11/2026.
Agents & tool use
| Benchmark | Meta Muse Ember | Claude Haiku 5.5 | GPT-6 Luna | Muse Spark 1.3 |
|---|---|---|---|---|
| AutomationBench-AAstrict pass@1; max reasoning | 37.8% | 13.9% | 20.2% | n/a |
| Agents' Last Exam (Linux subset)max reasoning; common internal agent loop | 61.6% | 57% | 57.1% | n/a |
| Personal Superintelligence Indexinternal composite of eight evaluations; max reasoning | 46.8% | 36.1% | 37% | n/a |
| OSWorld 2.1 (partial credit)108 workflows; common internal framework; max reasoning | 69.5% | 68.9% | 50.5% | n/a |
| OSWorld 2.1 (binary)strict full-completion rate; max reasoning | 33.8% | 27.8% | 17% | n/a |
Coding
| Benchmark | Meta Muse Ember | Claude Haiku 5.5 | GPT-6 Luna | Muse Spark 1.3 |
|---|---|---|---|---|
| Terminal-Bench 4.0best published harness: mini-swe-agent for Ember, Claude Code for Haiku 5.5, Codex for GPT-6 Luna; max reasoning | 40.9% | 39.2% | 16.4% | n/a |
| Coding Agent Index-AAArtificial Analysis; Muse Code for Ember, Claude Code for Haiku 5.5, Codex for Luna | 46 points | 37 | 31 | n/a |
| FrontierCode 2.0-AAArtificial Analysis private 132-task set; max reasoning | 27.2% | 34.9% | 25.2% | n/a |
Multimodal
| Benchmark | Meta Muse Ember | Claude Haiku 5.5 | GPT-6 Luna | Muse Spark 1.3 |
|---|---|---|---|---|
| GDP.pdf-AAall-pass share of answers meeting every rubric criterion; Artificial Analysis | 22.8% | 20.8% | 22.8% | n/a |
| PerceptionBench3,000 questions, with tools; pass@1 | 80.6% | 78.6% | 62.8% | n/a |
Other
| Benchmark | Meta Muse Ember | Claude Haiku 5.5 | GPT-6 Luna | Muse Spark 1.3 |
|---|---|---|---|---|
| Artificial Analysis Intelligence Indexmax reasoning; v5 | 36 points | 36 | 28 | 37 |
What did safety testing find for Meta Muse Ember?
Safety level: Meta Superintelligence Scaling Framework: below the framework's high-risk thresholds in the categories assessed (chemical and biological, cybersecurity, loss of control); no named risk level published
- Meta states it was evaluated under the Meta Superintelligence Scaling Framework for chemical and biological, cybersecurity and loss-of-control risks and remained below the high-risk thresholds.
- Received the same core safety training and testing as the larger Muse models.
- Behavioural evaluations covered respecting user intent, handling sensitive information, resisting hidden instructions in web pages, documents and tool outputs, and seeking human input before high-impact actions.
- Muse Ember is the first model trained using Muse Wonder Full.
- The blog gives no numeric safety scores; the published evaluation report covers capability methodology only.
Independent figures: Artificial Analysis, refreshed daily on our AI model benchmarks page. Full analysis: our Meta Muse Ember deep dive.
Meta Model Timeline
1M tokens input; up to 512K output tokens context
1M tokens at inference context
1M tokens at inference context
Specifications
Related Tools
Claude Opus 5.5
Claude Opus 5.5, launched 22 September 2026, is Anthropic's current Opus-line flagship, replacing Opus 5. It performs at roughly the level of Claude Fable 5.1 on most work while costing 40% less to run than Opus 5, at $4 input / $20 output per million tokens (both 20% down, cache reads cut 60% to $0.20). It leads on Terminal-Bench 4.0, GDPval-AA v2.1 and Humanity's Last Exam, and is the first Opus model to ship with Fable-5.1-class cybersecurity, biology and distillation safeguards.
Claude Sonnet 5.5
Claude Sonnet 5.5, released 28 September 2026, is the second model in Anthropic's Claude 5.5 family and the faster, lower-cost complement to Claude Opus 5.5. It keeps Sonnet 5's $2 input / $10 output per million tokens ($0.20 cache reads) while generating output over 30% faster, and scores 70.6% on Terminal-Bench 4.0, above Opus 5.5's 66.4%. It has a 1M-token context window, 128K max output, adaptive thinking, and is the first Sonnet with cyber and anti-distillation classifiers.
GPT-6.1 Sol
GPT-6.1 Sol, launched at OpenAI DevDay on 29 September 2026, is OpenAI's mid-tier workhorse model. OpenAI says it delivers nearly the same intelligence as GPT-6 Astra at one-fifth of Astra's token price: $2 input / $10 output per million tokens with cached input cut to $0.10. It scores 75.2% on DeepSWE v1.1 (high effort) and 71.4% on OSWorld 2.0, has a 1.05M-token context window and 128K max output, and is available in the API plus ChatGPT Work and Codex.
