CRUXEval-O (output prediction) — CRUXEval output-prediction split; DeepSeek column name for CRUXEval-O
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | Kimi K2 Base | Moonshot | 83.5 | 4 | 2026-06-13 |
| 2 | MiMo V2 Flash Base | Xiaomi | 79.1 | 2 | 2026-06-13 |
| 3 | DeepSeek-V3.1-Base | DeepSeek | 76.4 | 2 | 2026-06-13 |
| 4 | DeepSeek-V3.2-Exp-Base | DeepSeek | 74.9 | 2 | 2026-06-13 |
| 5 | DeepSeek-V3-Base | DeepSeek | 71 | 2 | 2026-06-05 |
| 6 | DeepSeek-V3 | DeepSeek | 69.8 | 1 | 2026-05-01 |
| 7 | Hy3 preview-Base | Tencent | 68.4 | 2 | 2026-06-05 |
| 8 | GLM-4.5-Base | Z.ai | 67.8 | 2 | 2026-06-05 |
| 9 | Qwen2.5 7B Base | Alibaba | 63.4 | 1 | 2026-05-03 |
| 10 | Stable-DiffCoder-8B-Base | ByteDance | 60 | 1 | 2026-06-05 |
| 11 | Qwen2.5 Instruct 72B | Alibaba | 59.9 | 1 | 2026-05-01 |
| 12 | Llama 3.1 Instruct 405B | Meta | 59.9 | 1 | 2026-05-01 |
| 13 | Qwen2.5 Coder Instruct 7B | Alibaba | 56 | 1 | 2026-06-05 |
| 14 | Granite 4.1 30B | IBM | 55.8 | 1 | 2026-06-15 |
| 15 | Qwen3 4B Base Dense | Alibaba | 50.7 | 1 | 2026-05-03 |
| 16 | DeepSeek-V2 | DeepSeek | 49.8 | 1 | 2026-05-01 |
| 17 | Granite 4.1 8B | IBM | 47.6 | 1 | 2026-06-15 |
| 18 | DeepSeek-Coder-6.7B-Base | DeepSeek | 41 | 1 | 2026-06-05 |
| 19 | Qwen2.5 3B Base | Alibaba | 26.1 | 1 | 2026-05-03 |
| 20 | Gemma 3 1B Base Dense | 3.7 | 1 | 2026-05-03 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.