OSWorld-Verified Results (All)
OSWorld-Verified Results (All) is run by XLANG Lab (OSWorld team). It has ranked 110 models, of which the catalogue holds 19, scoring from 62.9 to 90.2 on Success Rate (Avg±Std).
measured by XLANG Lab (OSWorld team) · the board itself
| Place | Model | Metric | Score |
|---|---|---|---|
| 1stof 110 | Intelligence Indeed Agent | Success Rate (Avg±Std) | 90.2 |
| 2ndof 110 | Claude Fable 5 | Success Rate (Avg±Std) (1m) | 86 |
| 3rdof 110 | Claude Opus 4.7 | Success Rate (Avg±Std) | 83.6 |
| 4thof 110 | Claude Opus 5 | Success Rate (Avg±Std) (1m) | 83.4 |
| 5thof 110 | Coasty | Success Rate (Avg±Std) | 82.8 |
| 6thof 110 | Claude Sonnet 4.6 | Success Rate (Avg±Std) | 81.5 |
| 7thof 110 | Muse Spark 1.1 | Success Rate (Avg±Std) | 80.7 |
| 8thof 110 | Holo3-35B-A3B | Success Rate (Avg±Std) | 80.4 |
| 9thof 110 | Gemini 3.1 Pro Preview | Success Rate (Avg±Std) | 78.3 |
| 10thof 110 | Claude Opus 4.5 | Success Rate (Avg±Std) | 76.3 |
| 11thof 110 | MiniMax M3 | Success Rate (Avg±Std) | 75.2 |
| 13thof 110 | Qwen 3.7 Plus | Success Rate (Avg±Std) | 73.3 |
| 14thof 110 | Kimi K2.6 | Success Rate (Avg±Std) | 73.1 |
| 15thof 110 | Agent S3 | Success Rate (Avg±Std) (n=10) | 72.6 |
| 19thof 110 | Agent S3 | Success Rate (Avg±Std) (n=1) | 66 |
| 20thof 110 | GPT-5 | Success Rate (Avg±Std) | 65.8 |
| 22ndof 110 | GBOX | Success Rate (Avg±Std) | 64.2 |
| 24thof 110 | Kimi K2.5 | Success Rate (Avg±Std) | 63.3 |
| 25thof 110 | Claude Sonnet 4.5 | Success Rate (Avg±Std) | 62.9 |
Read from the board on 2026-08-25