UI-TARS-2 ranks first on Windows Agent Arena.
ByteDance Seed Sep 2025 Closed weights Mid-size
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SWE-bench Verified | 68.7% | #88 of 155 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Online-Mind2Web | 88.2% | #2 of 10 | Lab model card | |
| Windows Agent Arena | 50.6% | #1 of 5 | Lab model card | |
| Terminal-Bench | 45.3% | #5 of 22 | Lab model card | |
| AndroidWorld | 73.3% | #7 of 23 | Lab model card | |
| OSWorld | 47.5% | #4 of 13 | Lab model card | |
| OSWorld-Verified | 53.1% | #17 of 33 | Official leaderboard | |
| BrowseComp-ZH | 50.5% | #15 of 19 | Lab model card | |
| BrowseComp | 29.6% | #78 of 81 | Lab model card | |
| UI-TARS 15 Games Collection | 59.8 | #1 of 1 | Lab model card |
The Self-Reported Index fits one score to the 6 capability results ByteDance Seed reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 135.4 to 142.9. How we compute it.
The benchmarks the lab chose to headline, with the scores it gave.