GUI Automation on OSWorld
86.8Overall Success RateAgentic Programming (Ours)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Agentic Programming (Ours)Type=Agentic framework, Max Steps=152026.06 | 86.8 | 93.5 | 80 | 100 | |
| Holo3-35B-A3BType=Specialized model, Max Steps=1002026.06 | 80.4 | 78.3 | 62.9 | 95.8 | |
| OpenAPA w/ gemini-3.1-proType=Agentic framework, Max Steps=1002026.06 | 78.3 | 71.7 | 65.7 | 83.3 | |
| VLAA-GUI w/ Opus 4.5Type=Agentic framework, Max Steps=1002026.06 | 76.3 | 66.6 | 61.1 | 91.7 | |
| HIPPO Agent w/ Opus 4.5Type=Agentic framework, Max Steps=1002026.06 | 74.5 | 60.4 | 64.3 | 87.5 | |
| Kimi K2.6Type=General model, Max Steps=1002026.06 | 73.1 | 76.7 | 55 | 79.2 | |
| Agent S3 w/ Opus 4.5 + GPT-5 bBoN (N=10)Type=Agentic framework, Max Steps=1002026.06 | 72.6 | 63 | 63.9 | 79.2 | |
| claude-sonnet-4-6Type=General model, Max Steps=1002026.06 | 72.1 | 78.5 | 60.2 | 91.7 | |
| Agent S3 w/ GPT-5 bBoN (N=10)Type=Agentic framework, Max Steps=1002026.06 | 69.9 | 69.5 | 58.7 | 79.2 | |
| UiPath Screen Agent w/ Opus 4.5Type=Agentic framework, Max Steps=1002026.06 | 67.1 | 69.5 | 53 | 70.8 | |
| Agent S3 [4] w/ Opus 4.5 bBoN (N=1)Type=Agentic framework, Max Steps=1002026.06 | 66 | 61.7 | 57.4 | 75 | |
| UI-MOPDCategory=Integration Strategies2026.07 | 38.2 | — | — | — | |
| Qwen3-VL-235B-A22B-ThinkingCategory=General Models2026.07 | 38.1 | — | — | — | |
| Model Merge (TIES Merging)Category=Integration Strategies2026.07 | 36.8 | — | — | — | |
| Model Merge (Weight Averaging)Category=Integration Strategies2026.07 | 36.5 | — | — | — | |
| Mixed-SFTCategory=Integration Strategies2026.07 | 35 | — | — | — | |
| GUI-Owl-7BCategory=GUI Models (Multi-Platform)2026.07 | 34.9 | — | — | — | |
| OpenCUA-32BCategory=GUI Models (Single-Platform)2026.07 | 34.8 | — | — | — | |
| SeedVL-1.5Category=General Models2026.07 | 34.1 | — | — | — | |
| Qwen3-VL-8B-InstructCategory=General Models2026.07 | 33.9 | — | — | — | |
| Qwen3-VL-8B-ThinkingCategory=General Models2026.07 | 33.9 | — | — | — | |
| Qwen3-VL-32B-InstructCategory=General Models2026.07 | 32.6 | — | — | — | |
| GELab-Zero-4BCategory=GUI Models (Multi-Platform)2026.07 | 31.9 | — | — | — | |
| Qwen3-VL-235B-A22B-InstructCategory=General Models2026.07 | 31.6 | — | — | — | |
| OpenAI CUA o3Category=GUI Models (Single-Platform)2026.07 | 31.3 | — | — | — | |
| OpenCUA-7BCategory=GUI Models (Single-Platform)2026.07 | 28.2 | — | — | — | |
| UI-TARS-1.5-7BCategory=GUI Models (Multi-Platform)2026.07 | 27.4 | — | — | — | |
| UI-TARS-72B-DPOCategory=GUI Models (Multi-Platform)2026.07 | 27.1 | — | — | — |