CRUXEval-I (input prediction) — CRUXEval input-prediction split; DeepSeek column name for what card labels CRUXEval-I
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | Kimi K2 Base | Moonshot | 74 | 4 | 2026-06-13 |
| 2 | Hy3 preview-Base | Tencent | 71.2 | 2 | 2026-06-05 |
| 3 | GLM-4.5-Base | Z.ai | 68.5 | 2 | 2026-06-05 |
| 4 | DeepSeek-V3-Base | DeepSeek | 67.8 | 2 | 2026-06-05 |
| 5 | MiMo V2 Flash Base | Xiaomi | 67.5 | 2 | 2026-06-13 |
| 6 | DeepSeek-V3 | DeepSeek | 67.3 | 1 | 2026-05-01 |
| 7 | DeepSeek-V3.2-Exp-Base | DeepSeek | 63.9 | 2 | 2026-06-13 |
| 8 | DeepSeek-V3.1-Base | DeepSeek | 62.1 | 2 | 2026-06-13 |
| 9 | Qwen2.5 Instruct 72B | Alibaba | 59.1 | 1 | 2026-05-01 |
| 10 | Llama 3.1 Instruct 405B | Meta | 58.5 | 1 | 2026-05-01 |
| 11 | DeepSeek-V2 | DeepSeek | 52.5 | 1 | 2026-05-01 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.