End-to-End Environment Interaction on AndroidWorld (test)
80.2Pass@1MobileRL-9B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MobileRL-9B2025.12 | 80.2 | — | |
| UI-TARS-22025.12 | 73.3 | — | |
| Mobile-Agent-v32025.12 | 73.3 | — | |
| Gemini-2.5-Computer Use2025.12 | 69.7 | — | |
| Step-GUI-8BEval Protocol=Pass@12025.12 | 67.7 | — | |
| GUI-OwlSize=7B2025.12 | 66.4 | — | |
| UI-TARS-1.52025.12 | 64.2 | — | |
| Step-GUI-4BEval Protocol=Pass@12025.12 | 63.9 | — | |
| SeedVL-1.52025.12 | 62.1 | — | |
| GLM-4.6V2025.12 | 57 | — | |
| Qwen3Size=30B-A3B, Variant=Instruct2025.12 | 54.3 | — | |
| Qwen3-VLSize=8B, Variant=Instruct2025.12 | 47.6 | — | |
| UI-TARSSize=72B, Variant=DPO2025.12 | 46.6 | — | |
| Qwen3-VLSize=4B, Variant=Instruct2025.12 | 45.3 | — | |
| Step-GUI-4BEval Protocol=Pass@32025.12 | — | 75.8 | |
| Step-GUI-8BEval Protocol=Pass@32025.12 | — | 80.2 |