GUI Agent Interaction on OSWorld
42.5Average AccuracyUI-TARS
Evaluation Results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| UI-TARSModel=UI-TARS-1.5-72B, Steps=100, Open Source=Close source2025.05 | 42.5 | — | — | — | — | — | — | — | — | — | — | |
| OpenAI CUAModel=GPT-4o, Steps=200, Open Source=Close source2025.05 | 38.1 | — | — | — | — | — | — | — | — | — | — | |
| UI-TARSModel=UI-TARS-1.5-72B, Steps=50, Open Source=Close source2025.05 | 38 | — | — | — | — | — | — | — | — | — | — | |
| InfantAgent-NextModel=Claude-3.7-Sonnet + UI-TARS-1.5-7B, Steps=50, Open Source=Open source2025.05 | 35.3 | — | — | — | — | — | — | — | — | — | — | |
| Agent S2Model=Claude-3.7-Sonnet, Steps=50, Open Source=Open source2025.05 | 34.5 | — | — | — | — | — | — | — | — | — | — | |
| OpenAI CUAModel=GPT-4o, Steps=50, Open Source=Close source2025.05 | 32.6 | — | — | — | — | — | — | — | — | — | — | |
| DynamicUIBackbone=Qwen3-vl-8B, Step=502026.04 | 28.4 | 34.8 | 69.6 | 17 | 19.6 | 26.1 | 13.1 | 62.5 | 40 | 17.7 | 43.5 | |
| Opencua-7BStep=502026.04 | 28.2 | 38.6 | 43.6 | 13.2 | 32.6 | 33.3 | 12.1 | 43.5 | 42.2 | 28.3 | 47.1 | |
| DynamicUIBackbone=UITARS-1.5-7B, Step=502026.04 | 28.2 | 32.8 | 57.7 | 8.5 | 39.3 | 41.3 | 8.6 | 29.8 | 50 | 27.5 | 54.4 | |
| Claude Computer UseModel=Claude-3.7-Sonnet, Steps=100, Open Source=Close source2025.05 | 28 | — | — | — | — | — | — | — | — | — | — | |
| UITARS-1.5-7BStep=502026.04 | 27.2 | 32.8 | 53.9 | 8.5 | 39.3 | 41.3 | 8.6 | 25.6 | 46.7 | 24.4 | 52.2 | |
| AGMem (Ours)Backbone Model=GPT-5.4-mini2026.06 | 27.2 | — | — | — | — | — | — | — | — | — | — | |
| UI-TARSModel=UI-TARS-1.5 7B, Steps=100, Open Source=Open source2025.05 | 26.9 | — | — | — | — | — | — | — | — | — | — | |
| Claude Computer UseModel=Claude-3.7-Sonnet, Steps=50, Open Source=Close source2025.05 | 26 | — | — | — | — | — | — | — | — | — | — | |
| UITARS-72b-dpoStep=502026.04 | 25.8 | 33.2 | 61.5 | 12.8 | 25.5 | 43.5 | 6.7 | 33.3 | 33.3 | 23.5 | 47.8 | |
| Qwen3-vl-8B*Step=502026.04 | 25.8 | 23.9 | 34.8 | 17 | 31.8 | 56.5 | 9.2 | 33.3 | 66.7 | 40.4 | 21.7 | |
| UI-TARSModel=UI-TARS-72B-DPO, Steps=50, Open Source=Open source2025.05 | 24.6 | — | — | — | — | — | — | — | — | — | — | |
| + visual memoryBackbone Model=GPT-5.4-mini2026.06 | 20.4 | — | — | — | — | — | — | — | — | — | — | |
| Opencua-a3bStep=502026.04 | 19.9 | 21.6 | 53.9 | 2.1 | 23.3 | 34.8 | 6.5 | 37.5 | 6.7 | 11.8 | 43.5 | |
| Qwen3-vl-4B*Step=502026.04 | 19.8 | 30.4 | 21.7 | 12.8 | 21.8 | 39.1 | 5.2 | 33.3 | 46.7 | 23 | 13 | |
| Vanilla agentBackbone Model=GPT-5.4-mini2026.06 | 18.3 | — | — | — | — | — | — | — | — | — | — | |
| o3Step=502026.04 | 17.2 | 21.7 | 38.5 | 8.5 | 4.3 | 21.7 | 11.8 | 37.5 | 20 | 17.6 | 21.7 | |
| AGUVISModel=AGUVIS 72B + GPT-4o, Open Source=Open source2025.05 | 17 | — | — | — | — | — | — | — | — | — | — | |
| Aria-UIModel=Aria-UI + GPT-4o, Open Source=Open source2025.05 | 15.1 | — | — | — | — | — | — | — | — | — | — | |
| OS-AtlasModel=OS-Atlas-Base-7B + GPT-4o, Open Source=Open source2025.05 | 14.6 | — | — | — | — | — | — | — | — | — | — | |
| SeeClickModel=SeeClick + GPT-4o, Open Source=Open source2025.05 | 9.21 | — | — | — | — | — | — | — | — | — | — | |
| Qwen2.5Model=Qwen2.5-vl-72B, Open Source=Open source2025.05 | 8.8 | — | — | — | — | — | — | — | — | — | — |