Muse Wonder (xHigh): benchmark results
Provider: Meta. Released 2026-10-12. Access: API.
Aggregate Rating 1918 ± 20, rank #32 of 2170 rated models, from 13 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LisanBench | 0.61 | Mean Path Length / Current Maximum | 97.6 |
| ProgramBench | 5.5 | Resolved (%) | 89.7 |
| DuelLab GameBench 2 - Reasoning Variants | 77.14 | GameBench2 normalized score (0–100) | 89.2 |
| ProgramBench Almost | 50.5 | Almost (%) | 86.2 |
| LLM2014 Logic 2026-10 | 59.64 | Median Score | 83.7 |
| CursorBench 4.0 | 48.2 | Score (%) | 82 |
| cua-speedrun - OSWorld 2.0 (52 Tasks) | 68.6 | Mean verifier score (%), partial credit over the task set | 77.8 |
| cua-speedrun - OSWorld (50 Tasks) | 89.62 | Mean verifier score (%), partial credit over the task set | 75.4 |
| Argo-Bench | 32.56 | Mean task score (0-100) over 210 enterprise data-science tas | 70 |
| Argo-Bench - Solved | 14.76 | Share of the 210 tasks solved (%), a task counting as solved | 66.7 |
| Agents on Rails - Atomic Task Score | 70.11 | Mean Task Score (%) | 41.7 |
| Agents on Rails - Feature Ticket Score | 35.15 | Mean Task Score (%) | 41.7 |
Interactive version: theaggregate.ai/model?slug=muse-wonder-xhigh · How It Works · Data refreshed daily, snapshot 2026-11-05.