| Benchmark | Claude 3.5 Sonnet | Nemotron 3 Super 120B A12B |
|---|---|---|
| AA Intelligence | 10 | 25.7 |
| AIME 2025 | 7.4 | 90.2 |
| Artificial Analysis Coding Index | 30.2 | 37.7 |
| GPQA Diamond | 67.2 | 82.7 |
| GSM8K | 96.9 | 90.7 |
| HLE | 3.9 | 22.8 |
| humaneval | 93.7 | 80.5 |
| LiveCodeBench | 32.8 | 82.1 |
| mmlu | 88.3 | 86 |
| MMLU-Pro | 78 | 83.8 |
| OpenBookQA | 97.2 | 48.6 |
| scicode | 36.6 | 42 |
| SWE-bench Verified | 50.8 | 53.7 |
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.