Agentic & Tool Use benchmark
OSWorld leaderboard
As of October 2026, Claude Sonnet 4.6 has the highest published OSWorld score on Noometry at 72.1%, out of 8 models with results.
Last verified
About OSWorld
Open-ended computer-use tasks in real desktop operating systems and applications.
- Category
- Agentic & Tool Use
- Introduced
- 2024
- Size
- 369 tasks
- Format
- Computer use
- Unit
- Percent (random guessing ≈ 0%)
- Official site
- os-world.github.io
Top 8 models
- Claude Sonnet 4.6 72.1%
- Claude Opus 4.5 66.3%
- Kimi K2.5 63.3%
- Claude Sonnet 4.5 62.9%
- Claude Sonnet 4 43.9%
- Claude 3.7 Sonnet 35.8%
- o3 23%
- Qwen2.5-VL 72B Instruct 5%
Sponsored placements are available on pages like this one. Advertise on Noometry
All results
| # | Model | Provider | Score | Setting | Source | Date |
|---|---|---|---|---|---|---|
| 1 | Claude Sonnet 4.6 | Anthropic | 72.1% | Epoch AI | ||
| 2 | Claude Opus 4.5 | Anthropic | 66.3% | Epoch AI | ||
| 3 | Kimi K2.5 | Moonshot AI | 63.3% | Epoch AI | ||
| 4 | Claude Sonnet 4.5 | Anthropic | 62.9% | Epoch AI | ||
| 5 | Claude Sonnet 4 | Anthropic | 43.9% | Epoch AI | ||
| 6 | Claude 3.7 Sonnet | Anthropic | 35.8% | Epoch AI | ||
| 7 | o3 | OpenAI | 23% | medium | Epoch AI | |
| 8 | Qwen2.5-VL 72B Instruct |  Alibaba (Qwen) | 5% | Epoch AI |
Compare the leaders
Frequently asked questions
What does OSWorld measure?
Open-ended computer-use tasks in real desktop operating systems and applications.
Which model has the highest OSWorld score?
As of October 2026, Claude Sonnet 4.6 has the highest published OSWorld score on Noometry at 72.1%, out of 8 models with results.
What is the best open-weight model on OSWorld?
Kimi K2.5 has the highest OSWorld accuracy among open-weight models at 63.3%, ranking 3 of 8 overall.