| Benchmark | Claude 3.5 Sonnet | DeepSeek-V4-Flash |
|---|---|---|
| AA Intelligence | 10 | 51.8 |
| Artificial Analysis Coding Index | 30.2 | 69.1 |
| Chinese SimpleQA (C-SimpleQA) | 56.8 | 78.9 |
| Codeforces | 717 | 3052 |
| GPQA Diamond | 67.2 | 90.8 |
| HLE | 3.9 | 38.6 |
| LiveCodeBench | 32.8 | 91.6 |
| MMLU-Pro | 78 | 86.4 |
| scicode | 36.6 | 49.9 |
| simplebench | 41.4 | 46.3 |
| SWE-bench Verified | 50.8 | 79 |
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.