Benchmarks / Agentic
OSWorld-Verified
Computer use on 361 verified Ubuntu desktop tasks (Chrome, LibreOffice, GIMP, VS Code…). Success rate of general models, best across step budgets.
OSWorld-Verified
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Fable 5 | max | 86.0% |
| 2 | Opus 5 | max | 83.4% |
| 3 | Muse Spark 1.1 | max | 80.7% |
| 4 | MiniMax M3 | max | 75.2% |
| 5 | Qwen 3.7 Plus | high | 73.3% |
| 6 | Kimi K2.6 | max | 73.1% |
| 7 | Sonnet 4.6 | max | 72.1% |
| 8 | Kimi K2.5 | max | 63.3% |
| 9 | Sonnet 4.5 | high | 62.9% |
| 10 | Qwen 3 | high | 41.6% |
| 11 | Sonnet 3.7 | medium | 35.8% |
| 12 | o3 | high | 23.0% |
| 13 | Qwen 2.5 | – | 3.9% |