| Benchmark | GPT-4o | Nemotron 3 Super 120B A12B |
|---|---|---|
| AA Agentic Index | 9.7 | 8.8 |
| AA Intelligence | 11.2 | 25.7 |
| AA-LCR | 55 | 60.3 |
| AA-Omniscience | -10.5 | -41.5 |
| AIME 2025 | 11.7 | 90.2 |
| Artificial Analysis Coding Index | 24.2 | 37.7 |
| critpt | 0 | 3.1 |
| gdpval | 0 | 9.9 |
| GPQA Diamond | 70.1 | 82.7 |
| GSM8K | 95.6 | 90.7 |
| HLE | 5.3 | 22.8 |
| humaneval | 90.6 | 80.5 |
| IFBench | 36 | 72.6 |
| LiveCodeBench | 38.3 | 82.1 |
| mmlu | 88.1 | 86 |
| mmlu_prox | 61.1 | 79.4 |
| MMLU-Pro | 74.7 | 83.8 |
| multichallenge | 40.3 | 55.2 |
| OmniScience Accuracy | 23.7 | 24.3 |
| OmniScience Non-Hallucination | 62.1 | 13 |
| OpenBookQA | 96.8 | 48.6 |
| scicode | 33.4 | 42 |
| SWE-bench Verified | 38.8 | 53.7 |
| Tau2 airline | 45.5 | 56.3 |
| Tau2 retail | 63.4 | 62.8 |
| Terminal-Bench Hard | 8.3 | 28.8 |
| τ²-Bench Telecom (AA run) | 28.9 | 67.8 |
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.