Muse Spark 1.3 (Medium): benchmark results
Provider: Meta. Released 2026-09-02. Access: API.
Aggregate Rating 1781 ± 22, rank #195 of 2274 rated models, from 23 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Coarena - Task Completion | 96.2 | Completion Rate (%) | 100 |
| ObviousBench | 98.61 | Answer pass³ (%) | 83.2 |
| DuelLab GameBench 2 - Reasoning Variants | 64.44 | GameBench2 normalized score (0–100) | 76.5 |
| Maze-Bench | 40.75 | Mean Score | 62.5 |
| Thrive Helpdesk - Difficult Tickets | 40.38 | Correct ticket resolution (%) | 41.1 |
| cua-speedrun - OSWorld (50 Tasks) | 77.54 | Mean verifier score (%), partial credit over the task set | 40 |
| Coarena | 987.2 | Elo (Bradley-Terry) | 36 |
| Argo-Bench - Solved | 9.52 | Share of the 210 tasks solved (%), a task counting as solved | 31 |
| Bug Hunt Bench - VS Code Extension | 5.3 | Planted Bugs Fixed (out of 45) | 24.4 |
| Web Clone Bench v1 - DocuSign | 0.4 | Full score on the DocuSign clone (%) | 24.4 |
| Web Clone Bench v1 - Gmail | 8.7 | Full score on the Gmail clone (%) | 23.1 |
| Web Clone Bench v1 - Slack | 4.9 | Full score on the Slack clone (%) | 23.1 |
Interactive version: theaggregate.ai/model?slug=muse-spark-1-3-medium · How It Works · Data refreshed daily, snapshot 2026-11-18.