Yutori

Last updated Sep 03, 2026

OSWorld 2.0 Leaderboard

Tracking reported computer-use agent performance on OSWorld 2.0, combining the official benchmark leaderboard with additional results found by our Scout.

Markdown for agents

500 steps selected

ModelOrganizationReasoningTool settingPartial score
Claude Fable 5.1AnthropicMax77.9%
GPT-6 AstraOpenAIMax72.6%
Claude Opus 5Anthropic70.6%
Muse Spark 1.3MetaMax66.9%
Claude Fable 5Anthropic66.1%
Navigator n21YutoriMediumBatched Tool65.2%
GPT-5.6 SolOpenAIMaxBatch Tool62.7%
Gemini 3.8 FlashGoogleBatched Tool59.0%
Kimi K3Moonshot AIMax58.3%
Claude Opus 4.8AnthropicMaxBatched Tool54.8%
Gemini 3.7 FlashGoogleBatched Tool50.6%
GPT-5.6 TerraOpenAIMax50.2%
GPT-5.5OpenAIXHighBatch Tool49.5%
Claude Opus 4.7AnthropicMaxStandard49.1%
Muse Spark 1.1MetaXHighBatched Tool47.3%
GPT-5.6 LunaOpenAI45.6%
Claude Sonnet 4.6AnthropicMaxStandard41.5%
Gemini 3.6 FlashGoogle33.8%
Gemini 3.1 ProGoogleHighBatched Tool30.6%2
MiniMax M3MiniMaxEnabledStandard22.3%
Kimi 2.6Moonshot AIEnabledStandard22.1%
Qwen 3.7-PlusAlibabaThinkingStandard21.5%

1 Navigator n2’s 500-step score is reported in the launch post. Its 150- and 300-step scores, reasoning level, and tool setting are first-party Yutori data supplied for this leaderboard.

2 Gemini 3.1 Pro result reported by Alibaba in the Qwen3.8 release post.

Frequently asked questions

Resources

Citation
@misc{yuan2026osworld2,
  title={OSWorld 2.0: Benchmarking Computer Use Agents on Long-Horizon Real-World Tasks},
  author={Mengqi Yuan and Zilong Zhou and Xinzhuang Xiong and Weiming Wu and Jiayang Sun and Jiamin Song and Kaiqian Cui and Bowen Wang and Haoyuan Wu and Yitong Li and Dunjie Lu and Haikong Lu and Qi Zhen and Xinyuan Wang and Jiaqi Deng and Yuhao Yang and Cheng Chen and Boyuan Zheng and Alex Su and Xiao Yu and Hao Zou and Saaket Agashe and Xing Han Lu and Manpreet Kaur and Zhengyang Qi and Vincent Sunn Chen and Frederic Sala and Dayiheng Liu and Junyang Lin and Zhou Yu and Yu Su and Siva Reddy and Xin Eric Wang and Peng Qi and Tianbao Xie and Tao Yu},
  year={2026},
  eprint={2606.29537},
  archivePrefix={arXiv},
  primaryClass={cs.AI},
}