Arena

Head-to-head on the public 60

Compare two models on individual tasks. Select runs, filter by pass/fail outcomes, then inspect trajectories side-by-side.

Attribution

The head-to-head Arena layout (model pickers, pass/fail outcome matrix, task list, side-by-side trajectories) follows the MobileWorld Arena interaction pattern. Please cite MobileWorld when referring to that design: 

@inproceedings{kong2025mobileworld,
  title={MobileWorld: Benchmarking Autonomous Mobile Agents in Agent-User Interactive, and MCP-Augmented Environments},
  author={Quyu Kong and Xu Zhang and Zhenyu Yang and Nolan Gao and Chen Liu and Panrong Tong and Chenglin Cai and Hanzhang Zhou and Jianan Zhang and Liangyu Chen and Zhidan Liu and Steven Hoi and Yue Wang},
  booktitle={Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (ACL)},
  year={2026},
  url={https://arxiv.org/abs/2512.19432},
}

Paper: arXiv:2512.19432 · Code: Tongyi-MAI/MobileWorld · Site: tongyi-mai.github.io/MobileWorld/arena. Pass/fail here follows each public run’s recorded outcome in the trajectory index (same source as the Tasks browser). Leaderboard Success Rate remains the manual-audit figure.