OSWorld 2.0 150-step Partial Score
Agents · 2026-06-25
OSWorld 2.0 official leaderboard partial score under a 150-step action budget over the 108-task v2 suite.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.5 | 46.7 |
| Opus 4.7 | 20.3 |
| Sonnet 4.6 | 20.0 |
| MiniMax M3 | 8.2 |
| Kimi K2.6 | 7.1 |