Windows Computer Use Automation on WindowsAgentArena V2
7.1LibreOffice Success RateUI-TARS-1.5-7B
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| UI-TARS-1.5-7B2025.05 | 7.1 | 34.7 | 23.1 | 45.8 | 21.1 | 7.6 | 16.7 | 21.3 | |
| PC Agent-Ebackbone=Qwen2.5-VL-72B2025.05 | 4.8 | 64.1 | 46.2 | 50 | 57.9 | 35.7 | 33.3 | 36 | |
| Claude 3.7 Sonnet2025.05 | 2.4 | 46.5 | 61.5 | 54.2 | 52.6 | 29 | 16.7 | 32.6 | |
| Claude 3.7 Sonnetthinking_mode=enabled2025.05 | 2.4 | 64.1 | 46.2 | 66.7 | 52.6 | 21.9 | 25 | 35.4 | |
| GPT-4o2025.05 | 0 | 5.9 | 0 | 8.3 | 0 | 0 | 0 | 2.1 | |
| Qwen2.5-VL-72B2025.05 | 0 | 34.7 | 15.4 | 20.8 | 26.3 | 7.6 | 16.7 | 14.9 | |
| UI-TARS-72B-DPO2025.05 | 0 | 40.6 | 38.5 | 58.3 | 36.8 | 7.6 | 25 | 26.2 |