Across 13 shared benchmarks, Nemotron 3 Super 120B A12B scores higher on 6 and NVIDIA-Nemotron-3-Ultra-550B-A55B-Base-BF16 on 6, with 1 level. The widest gap is GPQA Diamond, where Nemotron 3 Super 120B A12B scores 82.7 against 50.
| Benchmark | Nemotron 3 Super 120B A12B | NVIDIA-Nemotron-3-Ultra-550B-A55B-Base-BF16 |
|---|---|---|
| AGIEval-En | 77.9 | 78.7 |
| arc_challenge | 96.1 | 97.3 |
| GPQA Diamond | 82.7 | 50 |
| GSM8K | 90.7 | 88.1 |
| hellaswag | 89 | 90.5 |
| humaneval | 80.5 | 83.8 |
| mmlu | 86 | 89.1 |
| MMLU-Pro | 83.8 | 79.1 |
| OpenBookQA | 48.6 | 48.6 |
| RULER 1M | 93.9 | 76.8 |
| RULER 256K | 96.7 | 86.2 |
| RULER 512K | 95.7 | 84.5 |
| winogrande | 78.9 | 79.3 |
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.