GUI Automation on OSWorld overall
86.8Overall Success RateLLM-as-Code w/ Claude Sonnet 4.6
Evaluation Results
| Method | Links | |
|---|---|---|
| LLM-as-Code w/ Claude Sonnet 4.6Max Steps=152026.06 | 86.8 | |
| Holo3-35B-A3BMax Steps=1002026.06 | 80.4 | |
| OpenAPA w/ Gemini-3.1-proMax Steps=1002026.06 | 78.3 | |
| Claude Sonnet 4.6Max Steps=1002026.06 | 72.1 |