Computer Use on OSWorld
75OS Success RateEvoCUA-32B
Evaluation Results
| Method | Links | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| EvoCUA-32B2026.05 | 75 | 51.06 | 52.98 | — | — | — | — | 65.22 | — | — | 61.07 | — | — | — | — | |
| R-WoMBackbone=Claude-4.5-Sonnet2025.10 | 67.84 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| WoMBackbone=Claude-4.5-Sonnet2025.10 | 62.09 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RAGBackbone=Claude-4.5-Sonnet2025.10 | 60.43 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VanillaBackbone=Claude-4.5-Sonnet2025.10 | 59.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| R-WoMBackbone=Claude-4-Sonnet2025.10 | 56.73 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RAGBackbone=Claude-4-Sonnet2025.10 | 50.82 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| WoMBackbone=Claude-4-Sonnet2025.10 | 49.65 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| R-WoMBackbone Model=Claude-4.5-Sonnet, Setting=tutorial-scarce2025.10 | 49.29 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VanillaBackbone=Claude-4-Sonnet2025.10 | 48.24 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| WebDreamerBackbone Model=Claude-4.5-Sonnet, Setting=tutorial-scarce2025.10 | 46.35 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RAGBackbone Model=Claude-4.5-Sonnet, Setting=tutorial-scarce2025.10 | 46.11 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VanillaBackbone Model=Claude-4.5-Sonnet, Setting=tutorial-scarce2025.10 | 45.83 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| No DefenseAgent=Claude Sonnet 4.52026.02 | 42.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GPT-4o + AllyInput Modalities=Accessibility tree (A11y) and screenshot, Planner=GPT-4o, Grounding=Ally2024.12 | 41.67 | 4.26 | 6.81 | 8.7 | 9.5 | 6.67 | 15.22 | 30.43 | 0 | 7.46 | 11.21 | — | — | — | — | |
| DEACTIONAgent=Claude Sonnet 4.52026.02 | 40.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DEACTIONAgent=OpenCUA-72B2026.02 | 39.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| R-WoMBackbone Model=Claude-4-Sonnet, Setting=tutorial-scarce2025.10 | 39.28 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| No DefenseAgent=OpenCUA-72B2026.02 | 39 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DSPAgent=OpenCUA-72B2026.02 | 38.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RAGBackbone Model=Claude-4-Sonnet, Setting=tutorial-scarce2025.10 | 36.93 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Task ShieldAgent=OpenCUA-72B2026.02 | 36.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InferActAgent=OpenCUA-72B2026.02 | 36.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VanillaBackbone Model=Claude-4-Sonnet, Setting=tutorial-scarce2025.10 | 35.82 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| R-WoMBackbone Model=Claude-3.7-Sonnet, Setting=tutorial-scarce2025.10 | 35.71 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PromptArmorAgent=OpenCUA-72B2026.02 | 35.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| WebDreamerBackbone Model=Claude-4-Sonnet, Setting=tutorial-scarce2025.10 | 34.43 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RAGBackbone Model=Claude-3.7-Sonnet, Setting=tutorial-scarce2025.10 | 33.36 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| UI-TARS-1.5-7B + LEARNWEAK-GENframework=LEARNWEAK-GEN2026.05 | 33.33 | 8.51 | 22.7 | — | — | — | — | 40.57 | — | — | 26.28 | — | — | — | — | |
| VanillaBackbone Model=Claude-3.7-Sonnet, Setting=tutorial-scarce2025.10 | 32.25 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| WebDreamerBackbone Model=Claude-3.7-Sonnet, Setting=tutorial-scarce2025.10 | 30.86 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DEACTIONAgent=OpenAI CUA2026.02 | 30.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| No DefenseAgent=OpenAI CUA2026.02 | 26 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GPT-4o + Aria-UI_THInput Modalities=Pure-vision (screenshots only), Planner=GPT-4o, Grounding=Aria-UI_TH2024.12 | 25 | 4.26 | 15.32 | 8.7 | 30.06 | 26.67 | 23.8 | 21.74 | 19.23 | 8.55 | 15.15 | — | — | — | — | |
| GPT-4o + SoMInput Modalities=Accessibility tree (A11y) and screenshot, Planner=GPT-4o, Grounding=SoM2024.12 | 20.83 | 0 | 6.77 | 4.35 | 6.53 | 0 | 4.35 | 4.35 | 0 | 3.6 | 4.59 | — | — | — | — | |
| UI-TARS-1.5-7B2026.05 | 16.67 | 7.09 | 21.98 | — | — | — | — | 30.43 | — | — | 19.04 | — | — | — | — | |
| Claude2025.02 | 14.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Aguvis-72B2025.02 | 10.26 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen2.5-VL-72B2025.02 | 8.83 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GPT-4oInput Modalities=Pure-vision (screenshots only)2024.12 | 8.33 | 0 | 6.77 | 4.35 | 16.1 | 0 | 4.35 | 4.35 | 3.85 | 5.58 | 5.03 | — | — | — | — | |
| GPT-4o2025.02 | 5.03 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Gemini 2.02025.02 | 4.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CogAgent + SoMInput Modalities=Accessibility tree (A11y) and screenshot, Planner=CogAgent, Grounding=SoM2024.12 | 4.17 | 2.17 | 0 | 4.34 | 6.53 | 0 | 2.17 | 0 | 0 | 0 | 0.99 | — | — | — | — | |
| CogAgentInput Modalities=Pure-vision (screenshots only)2024.12 | 4.17 | 2.17 | 0 | 4.35 | 6.53 | 0 | 2.17 | 0 | 0 | 0.1 | 1.11 | — | — | — | — | |
| Qwen2-VL-72B2025.02 | 2.42 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Claude-Sonnet-4.52026.03 | — | — | — | — | — | — | — | — | — | — | — | 62.9 | — | — | — | |
| Gemini-2.5-Pro2026.03 | — | — | — | — | — | — | — | — | — | — | — | 13.3 | — | — | — | |
| GPT-O3-CUA2026.03 | — | — | — | — | — | — | — | — | — | — | — | 38.1 | — | — | — | |
| MACUSetup=Multi-Agent, Manager=Qwen2.5-72B, Subagent backbone=Qwen2.5-72B2026.06 | — | — | — | — | — | — | — | — | — | — | 44.9 | — | 1.1 | 14.2 | -12.4 | |
| MACUSetup=Multi-Agent, Manager=Opus 3.5, Subagent backbone=Qwen2.5-72B2026.06 | — | — | — | — | — | — | — | — | — | — | 48.5 | — | 4.7 | 21.4 | -5.2 | |
| Seed1.5-VL2026.03 | — | — | — | — | — | — | — | — | — | — | — | 36.7 | — | — | — | |
| Seed1.82026.03 | — | — | — | — | — | — | — | — | — | — | — | 61.9 | — | — | — | |
| Single-AgentSetup=Single-Agent, Manager=—, Subagent backbone=Qwen2.5-72B2026.06 | — | — | — | — | — | — | — | — | — | — | 43.8 | — | — | 26.6 | — |