CMath — CMath with explicit EM metric; kept separate from unlabeled CMath (different score range/source)
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | DeepSeek-V4-Flash-Base | DeepSeek | 93.6 | 4 | 2026-06-27 |
| 2 | DeepSeek-V3.2-Base | DeepSeek | 92.6 | 4 | 2026-06-27 |
| 3 | DeepSeek-V4-Pro-Base | DeepSeek | 90.9 | 4 | 2026-06-27 |
| 4 | DeepSeek-V3 | DeepSeek | 90.7 | 1 | 2026-05-01 |
| 5 | Qwen2.5 Instruct 72B | Alibaba | 84.5 | 1 | 2026-05-01 |
| 6 | DeepSeek-V2 | DeepSeek | 78.7 | 1 | 2026-05-01 |
| 7 | Llama 3.1 Instruct 405B | Meta | 77.3 | 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.