| Rank | Model | Provider | Score |
|---|---|---|---|
| 1 | Claude Opus 5.5 | Anthropic | 67.7% |
| 2 | Claude Fable 5.1 | Anthropic | 65.6% |
| 3 | DeepSeek V4.1 Flash | DeepSeek | 63.9% |
| 4 | Claude Opus 5 | Anthropic | 63.6% |
| 5 | GLM 5.3 | Zhipu AI | 62.5% |
| 6 | DeepSeek V4 Pro 0813 | DeepSeek | 60.0% |
| 7 | Kimi K3 | Moonshot AI | 59.8% |
| 8 | GPT 6 Astra | OpenAI | 57.2% |
| 9 | Qwen3.8 Max | Alibaba | 56.2% |
| 10 | GLM 5.3 Flash | Zhipu AI | 55.3% |
| 11 | DeepSeek V4 Flash 0731 | DeepSeek | 51.5% |
| 12 | Qwen3.6 Plus | Alibaba | 50.6% |
Figures are taken from public model cards, technical reports and vendor announcements, and are approximate where sources disagree. Results from different benchmark versions or tool-use settings never share a column: Terminal Bench 2.1, 3.0 and DeepSWE v1.1 are distinct benchmarks, and HLE with tools is separate from HLE without tools. Only models with a published result are listed; the score column is shaded relative to the best and worst result on this benchmark.