| Benchmark | Nemotron 3 Super 120B A12B | o3 |
|---|---|---|
| AA Agentic Index | 8.8 | 36.1 |
| AA Intelligence | 25.7 | 31.1 |
| AA-LCR | 60.3 | 73.3 |
| AA-Omniscience | -41.5 | -15.3 |
| AIME 2025 | 90.2 | 89.2 |
| Artificial Analysis Coding Index | 37.7 | 38.4 |
| browsecomp | 31.3 | 49.7 |
| critpt | 3.1 | 1.1 |
| gdpval | 9.9 | 12.8 |
| GPQA Diamond | 82.7 | 83.3 |
| HLE | 22.8 | 20.3 |
| IFBench | 72.6 | 71.4 |
| LiveCodeBench | 82.1 | 84.7 |
| MMLU-Pro | 83.8 | 85 |
| multichallenge | 55.2 | 60.4 |
| OmniScience Accuracy | 24.3 | 38.6 |
| OmniScience Non-Hallucination | 13 | 12.9 |
| scicode | 42 | 41 |
| SWE-bench Verified | 53.7 | 69.1 |
| Tau2 airline | 56.3 | 64.8 |
| Tau2 retail | 62.8 | 80.2 |
| Terminal-Bench Hard | 28.8 | 37.1 |
| τ²-Bench Telecom (AA run) | 67.8 | 80.7 |
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.