Computer Use on OSWorld (Verified)
75ScoreGPT-5.4
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-5.42026.05 | 75 | |
| Kimi-K2.62026.05 | 73.1 | |
| Claude-Sonnet-4.62026.05 | 72.5 | |
| Claude Opus 4.5Evaluation Source=External2026.02 | 66.3 | |
| Kimi K2.5Evaluation Source=External2026.02 | 63.3 | |
| Qwen-3.5-27B2026.05 | 56.2 | |
| GUI-OWL-1.5-8B2026.05 | 52.3 | |
| EvoCUA-8B2026.05 | 46.1 | |
| Qwen-3.5-9B2026.05 | 41.8 | |
| Qwen3-VL-235B-A22BEvaluation Source=External2026.02 | 38.1 | |
| Gemini 3 ProEvaluation Source=Internal2026.02 | 20.7 | |
| GPT-5.2 (xhigh)Evaluation Source=Internal2026.02 | 8.6 |