- OSWorld 47.5
- WAA (WindowsAgentArena) 50.6
- TB (Terminal Bench) with GUI-SDK, 75 of 80 tasks 45.3
- SB (SWE-Bench) with GUI-SDK 68.7
- AndroidWorld 73.3
- Online-Mind2web 88.2
- BC-zh 32.1 GUI-only (50.5 with GUI-SDK) 50.5
- BC-en 7.0 GUI-only (29.6 with GUI-SDK) 29.6
- 15 Games Collection, Mean Normalized Score (UI-TARS-2-RL) 59.8
UI-TARS-2
UI-TARS-2 ranks first on 2 of the 9 benchmarks it has scores on.
10
Benchmarks with a score
2
Ranked first, out of five or more models
3
In the top three
ByteDance Seed Sep 2025 UI-TARS-2
Closed weights Mid-size
No Epoch Capabilities Index score yet
- BrowseComp-ZH 50.5% #13 /17
- Windows Agent Arena 50.6% #1 /5
- BrowseComp 29.6% #74 /77
Where it ranks
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.
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SWE-bench Verified | 68.7% | #85 of 150 | Model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Online-Mind2Web | 88.2% | #1 of 8 | Model card | |
| Windows Agent Arena | 50.6% | #1 of 5 | Model card | |
| Terminal-Bench | 45.3% | #5 of 22 | Model card | |
| OSWorld | 47.5% | #3 of 12 | Model card | |
| AndroidWorld | 73.3% | #6 of 18 | Model card | |
| OSWorld-Verified | 53.1% | #17 of 33 | Official leaderboard | |
| BrowseComp-ZH | 50.5% | #13 of 17 | Model card | |
| BrowseComp | 29.6% | #74 of 77 | Model card | |
| UI-TARS 15 Games Collection | 59.8 | #1 of 1 | Model card |
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.