Meta
Muse Ember
Released · meta.ai
Context window
| Tokens | Context window | Output limit |
|---|
| Default | 1M | 512K |
Context history
Benchmark results
| Benchmark | Score |
|---|
| Terminal-Bench 4.0Run by Artificial Analysis · mini-swe-agent harness; max reasoning effort | 40.9% |
| AutomationBench 1.0.6AutomationBench-AA · Run by Artificial Analysis · max reasoning effort; pass@1 | 37.8% |
| OSWorld 2.1Full set, partial score · max reasoning effort; mean per-task partial credit; 108 tasks; Meta's own framework | 69.5% |
| OSWorld 2.1Full set, strict pass rate · max reasoning effort; share of tasks with full credit; 108 tasks; Meta's own framework | 33.8% |
Sources
Reported by Meta
All models