| Benchmark | Claude 3.5 Sonnet | DeepSeek-V3.2 |
|---|---|---|
| AA Intelligence | 10 | 32.8 |
| AIME 2025 | 7.4 | 93.1 |
| Artificial Analysis Coding Index | 30.2 | 44.2 |
| frontiermath_tier_4 | 0 | 2.1 |
| GPQA Diamond | 67.2 | 84 |
| HLE | 3.9 | 40.8 |
| hmmt_feb_2025 | 1.7 | 92.5 |
| LiveCodeBench | 32.8 | 86 |
| LiveCodeBench v6 | 37.2 | 83.3 |
| longbench_v2 | 41 | 59.8 |
| mmlu_redux | 88.9 | 93.7 |
| MMLU-Pro | 78 | 86 |
| scicode | 36.6 | 39 |
| SWE-bench Verified | 50.8 | 73.1 |
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.