Nemotron 3 Nano 30B-A3B Base.
NVIDIA Dec 2025 Open weights Mid-size
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| PIQA | 84.3% | #11 of 37 | Lab model card | |
| ARC-Challenge | 91.9% | #15 of 49 | Lab model card | |
| AGIEval | 68.3% | #14 of 40 | Lab model card | |
| WinoGrande | 79.6% | #21 of 47 | Lab model card | |
| HellaSwag | 85.6% | #26 of 53 | Lab model card | |
| RACE | 88% | #4 of 8 | Lab model card | |
| MGSM | 83% | #30 of 57 | Lab model card | |
| MMLU | 78.6% | #82 of 131 | Lab model card | |
| MMLU-Pro | 65% | #135 of 167 | Lab model card | |
| RULER | 75.4% | #36 of 42 | Lab model card | |
| OpenBookQA | 46.2% | #22 of 24 | Lab model card | |
| Global MMLU Lite 2.0 | 74.5% | #1 of 1 | Lab model card |
The Self-Reported Index fits one score to the 16 capability results NVIDIA reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 118.9 to 126.1. How we compute it.
Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.
Every result in the longer documents the lab published with it. 17 of them aren't in the launch post.