Benchmark
End-to-end tasks completed in a real terminal environment (v4.0 task set).
Score history →| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 5.5 | 66.4% |
| 2 | GPT-6 Astra | 57.9% |
| 3 | Gemini 4 Argon | 57.4% |
| 4 | Claude Opus 5 | 51.8% |
| 5 | DeepSeek V4.1-Flash 1017 | 42.3% |
| 6 | GPT 5.6-Sol | 39.9% |
| 7 | Grok 4.7 | 38% |
| 8 | GLM 5.3 | 37.9% |
| 9 | DeepSeek V4.1-Flash 0910 | 31.2% |
| 10 | Kimi K3 | 12.6% |
| 11 | DeepSeek V4-Pro 0813 | 12.4% |
| 12 | DeepSeek V4-Flash 0731 | 7% |