End-to-End Environment Interaction on OSWorld-Verified (test)
61.4Pass@1Claude-4.5-sonnet
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Claude-4.5-sonnet2025.12 | 61.4 | — | |
| UI-TARS-22025.12 | 47.5 | — | |
| Claude-4-sonnet2025.12 | 43.9 | — | |
| UI-TARS-1.52025.12 | 42.5 | — | |
| Step-GUI-8BEval Protocol=Pass@12025.12 | 40.2 | — | |
| Mobile-Agent-v32025.12 | 37.7 | — | |
| GLM-4.6V2025.12 | 37.2 | — | |
| GUI-OwlSize=7B2025.12 | 34.9 | — | |
| SeedVL-1.52025.12 | 34.1 | — | |
| Qwen3-VLSize=8B, Variant=Instruct2025.12 | 33.9 | — | |
| Step-GUI-4BEval Protocol=Pass@12025.12 | 31.9 | — | |
| Qwen3Size=30B-A3B, Variant=Instruct2025.12 | 30.3 | — | |
| UI-TARS-1.5Size=7B2025.12 | 27.4 | — | |
| Qwen3-VLSize=4B, Variant=Instruct2025.12 | 26.2 | — | |
| UI-TARSSize=72B, Variant=DPO2025.12 | 24 | — | |
| OpenAI CUA o32025.12 | 23 | — | |
| Step-GUI-4BEval Protocol=Pass@32025.12 | — | 40.4 | |
| Step-GUI-8BEval Protocol=Pass@32025.12 | — | 48.5 |