| Rank | Model | Provider | Score |
|---|---|---|---|
| 1 | GPT 5.6 Sol | OpenAI | 94.1% |
| 2 | Claude Opus 5 | Anthropic | 93.4% |
| 3 | Kimi K3 | Moonshot AI | 92.9% |
| 4 | Qwen3.8 Max | Alibaba | 92.6% |
| 5 | DeepSeek V4 Pro 0813 | DeepSeek | 92.4% |
| 6 | Qwen3.8 Flash Next | Alibaba | 91.7% |
| 7 | DeepSeek V4.1 Flash | DeepSeek | 90.9% |
| 8 | Qwen3.6 Plus | Alibaba | 90.4% |
| 9 | DeepSeek V4 Flash 0731 | DeepSeek | 89.9% |
| 10 | Qwen3.8 27B | Alibaba | 89.2% |
| 11 | GLM 5.3 | Zhipu AI | 88.1% |
| 12 | Qwen3.6 27B | Alibaba | 87.8% |
| 13 | Qwen3.6 35B A3B | Alibaba | 86.0% |
| 14 | o1 preview | OpenAI | 78.3% |
| 15 | Claude 3.5 Sonnet | Anthropic | 65.0% |
| 16 | o1 mini | OpenAI | 60.0% |
| 17 | GPT 4o | OpenAI | 53.6% |
| 18 | Llama 3.1 405B | Meta | 51.1% |
| 19 | Claude 3 Opus | Anthropic | 50.4% |
| 20 | Qwen2.5 72B | Alibaba | 49.0% |
| 21 | DeepSeek V2.5 | DeepSeek | 48.0% |
| 22 | Mistral Large 2 | Mistral AI | 48.0% |
| 23 | Gemini 1.5 Pro | 46.2% | |
| 24 | Llama 3.1 70B | Meta | 41.7% |
| 25 | GPT 4o mini | OpenAI | 40.2% |
| 26 | Gemini 1.5 Flash | 38.6% | |
| 27 | Llama 3.1 8B | Meta | 30.4% |
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.