All benchmarksIndependent evaluation · Epoch AI
OSWorld
Each model’s result as Epoch AI’s Benchmarking Hub publishes it, run by Epoch or collected from the source it cites. Facthory has not run them.
Data: Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks, licensed CC BY 4.0. Read on 16 Nov 2026.
| # | Model | Organization | Result | Evidence |
|---|---|---|---|---|
| 1 | Claude Sonnet 4.6 (no thinking) API access | Anthropic | 72.1% | Log or source |
| 2 | Claude Opus 4.5 (no thinking) API access | Anthropic | 66.3% | Log or source |
| 3 | Kimi K2.5 Open weights (unrestricted) | Moonshot | 63.3% | Log or source |
| 4 | Claude Sonnet 4.5 (no thinking) API access | Anthropic | 62.9% | Log or source |
| 5 | claude-sonnet-4-20250514 API access | Anthropic | 43.9% | Log or source |
| 6 | claude-3-7-sonnet-20250219 API access | Anthropic | 35.8% | Log or source |
| 7 | computer-use-preview-2025-03-11 | – | 31.3% | Log or source |
| 8 | o3 (medium) API access | OpenAI | 23% | Log or source |
| 9 | Qwen2.5-VL-72B-Instruct Open weights (unrestricted) | Alibaba | 5% | Log or source |
