Desktop Operating System Task Execution on OSWorld
20.3AUVUI-TARS-72B-DPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| UI-TARS-72B-DPOParameters=72B, Training=DPO2026.02 | 20.3 | 16.1 | |
| GPT-4o + ScaleCUA-7BPlanner=GPT-4o, Grounder=ScaleCUA-7B2026.02 | 15.8 | 36.8 | |
| UI-TARS-1.5-7BParameters=7B2026.02 | 15.4 | 21.3 | |
| Claude3.7-Sonnet-20250219Proprietary=true2026.02 | 8.4 | 1.4 | |
| Qwen2.5-VL-72B-InstructParameters=72B, Type=Instruct2026.02 | 5.9 | 4.6 |