Muse Ember: benchmark results
Meta's Muse Ember (muse-ember-1.0), released November 9, 2026, is a multimodal reasoning model for repeated coding subtasks, tool calls and document processing. Available through Meta Model API and Muse Code in Standard mode; accepts text, images, video, audio and PDF, with a 1,048,576-token context and up to 524,288 output tokens. Official release: https://research.meta.ai/blog/muse-ember; model documentation: https://dev.meta.ai/docs/muse-ember Provider: Meta. Released 2026-11-09. Access: API.
Aggregate Rating 1825 ± 11, rank #54 of 10870 rated models, from 130 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AutomationBench-AA - HR - Raw Objective Completion | 80.95 | Raw Objective Completion (%) | 98 |
| AA-Omniscience v1.1 Hallucination Rate | 17.34 | Hallucination Rate (%) | 97 |
| Design Arena (Website) | 1329 | Elo | 96.6 |
| Vals AI Harvey Legal Agent Bench v1.1 | 25 | Task Resolution (%) | 95.8 |
| AI Chess Leaderboard (Continuation) | 1507 | Elo | 95.1 |
| AutomationBench-AA - Operations - Raw Objective Completion | 96.51 | Raw Objective Completion (%) | 95.1 |
| AutomationBench-AA - Finance - Raw Objective Completion | 90.69 | Raw Objective Completion (%) | 94.9 |
| AutomationBench-AA - Marketing - Raw Objective Completion | 89.5 | Raw Objective Completion (%) | 94.9 |
| AA GDP.pdf - Manufacturing/Supply Chains - Criterion Pass Rate | 78.86 | Criterion Pass Rate (%) | 94.1 |
| AA GDP.pdf - Real Estate - Criterion Pass Rate | 76.42 | Criterion Pass Rate (%) | 94.1 |
| Design Arena (3D) | 1374 | Elo | 93.9 |
| Tinybird AI SQL Benchmark - Exactness | 56.98 | Result exactness vs human reference queries (0-100) | 93.9 |
Interactive version: theaggregate.ai/model?slug=muse-ember · How It Works · Data refreshed daily, snapshot 2026-11-12.