OSWorld
Rank #34 of 51 · 8 models scored · top score 72.1
unknown
Strongest on discrimination, weakest on contamination resistance. Caveat: the public test set leaves contamination on the table.
Weighted composite. Each signal is computed from the source-backed leaderboard below.
Does it still separate models?
How far from ceiling / clustered at the top?
Public vs held-out; training-leak risk.
Apples-to-apples, or mixed / vendor-optimized?
How old is the benchmark?
| # | Model | Score | Evidence | Measured |
|---|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 72.1 | unverified· optimized | 2026-02-17 |
| 2 | Claude Opus 4.5 | 66.3 | unverified· optimized | 2025-11-24 |
| 3 | Kimi K2.5 | 63.3 | unverified· optimized | 2026-02-02 |
| 4 | Claude Sonnet 4.5 | 62.9 | unverified· optimized | 2025-09-29 |
| 5 | Claude Sonnet 4 | 43.9 | unverified· optimized | 2025-05-22 |
| 6 | Claude 3.7 Sonnet | 35.8 | unverified· optimized | 2025-02-24 |
| 7 | o3 | 23 | unverified· optimized | 2024-12-20 |
| 8 | Qwen2.5-72B | 5 | unverified· optimized | 2024-09-19 |