Computer Agent Arena is an open-ended evaluation platform that tests AI agents in real computer-use scenarios—spanning desktop apps, websites, and more. This repository will soon host:
- Source code for the Arena platform
- Evaluation data (benchmarks, logs, results)
- Documentation on how to set up and extend the Arena
We are still preparing the code and data for an official release. Stay tuned for our first major update, coming soon.
- Platform: arena.xlang.ai
- Leaderboard (tentative): arena.xlang.ai/leaderboard
- Project Blog: arena.xlang.ai/blog/computer-agent-arena
We welcome contributions once the repository is populated with code. Please watch or star this repo for updates. We’ll share a detailed contribution guide soon.
If you find this project useful, please consider citing our project:
@misc{wang2025computer,
title={Computer Agent Arena: Compare & Test Computer Use Agents on Crowdsourced Real-World Tasks},
url={https://arena.xlang.ai}
author={Bowen Wang and Xinyuan Wang and Jiaqi Deng and Tianbao Xie and Ryan Li and Yanzhe Zhang and Gavin Li and Toh Jing Hua and Yu Su and Diyi Yang and Yi Zhang and Zhiguo Wang and Victor Zhong and Tao Yu},
year={2025},
}
© 2025 Computer Agent Arena Team