OSWorld 2.1 leaderboard
As of November 14, 2026, among the results listed here, Anthropic Claude Opus 5.5 has the highest reported OSWorld 2.1 score, 81.8% (partial), published by Anthropic, ahead of Google Gemini 4 Carbon at 81.5%. 9 models have a OSWorld 2.1 result in our records.
Scores as the model makers published them, not re-run by us; each row links to its announcement.
9 models ranked · 2 makers
| Rank | Model | Score | Setting | Reported by | Date |
|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 | 81.8% | partial | Anthropic | June 30, 2026 (model release) |
| 2 | Gemini 4 Carbon | 81.5% | Offline subset | October 28, 2026 · read from the maker's chart | |
| 3 | Claude Fable 5.1 | 80.7% | partial | Anthropic | July 24, 2026 (model release) |
| 4 | Claude Sonnet 5.5 | 80.1% | partial | Anthropic | June 30, 2026 (model release) |
| 5 | Gemini 4 Argon | 74.7% | Offline subset | October 28, 2026 · read from the maker's chart | |
| 6 | Claude Opus 5 | 74.0% | partial | Anthropic | July 24, 2026 (model release) |
| 7 | Claude Haiku 5.5 | 72.4% | Offline subset | Anthropic | October 7, 2026 (model release) |
| 8 | Claude Sonnet 5 | 57.0% | partial | Anthropic | June 30, 2026 (model release) |
| 9 | Claude Haiku 4.5 | 15.7% | Offline subset | Anthropic | October 7, 2026 (model release) |
How much to trust these numbers
- Source
- Reported by the model makers themselves. Not re-run by LLMBenchmark or anyone independent.
- Settings
- Settings differ between makers. Check the Setting column before comparing two scores.
- Spread
- 81.8% to 15.7% across 9 models.
- Freshness
- Newest result published October 28, 2026; oldest June 30, 2026.
- Charts
- Some values were read from charts in the announcements. Two independent readings had to match exactly.
What OSWorld 2.1 measures
OSWorld 2.1 scores reflect how well agents operate a real computer to complete long, multi-step tasks. Source: OSWorld-V2