OSWorld 2.0 Partial: leaderboard
Partial-score companion metric for OSWorld 2.0, measuring graded progress on the same long-horizon desktop workflows in addition to binary task completion.
Metric: Partial Score (%). Source: osworld-v2.xlang.ai. Saturation forecast: Around January 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 79.2 | #12 |
| 2 | GPT-5.6 Sol | 64.13 | #19 |
| 3 | Claude Opus 4.8 | 54.8 | #37 |
| 4 | GPT-5.5 | 49.5 | #31 |
| 5 | Claude Opus 4.7 | 49.1 | #49 |
| 6 | Claude Sonnet 4.6 | 41.5 | #90 |
| 7 | MiniMax-M3 | 22.3 | #136 |
| 8 | Kimi K2.6 | 22.1 | #104 |
| 9 | Qwen 3.7 Plus | 21.5 | #92 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=osworld-2-0-partial · How It Works · Data refreshed daily, snapshot 2026-10-20.