NVIDIA-Nemotron-3-Nano-30B-A3B-FP8: benchmark results
NVIDIA Nemotron 3 Nano 30B-A3B uses 23 Mamba-2 blocks, 23 MoE blocks and six attention blocks. Each MoE routes to six of 128 experts plus one shared expert; base, post-trained reasoning modes and checkpoint precision variants remain distinct. Provider: NVIDIA. Released 2025-12-15. Access: Open.
Aggregate Rating 1585 ± 25, rank #1181 of 10916 rated models, from 273 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| EuroEval Dutch NLU - CoNLL NL | 80.62 | Micro F1 excluding MISC (%) | 99.4 |
| EuroEval Polish Summarization - PSC | 25.74 | chrF3++ (%) | 99.3 |
| EuroEval Italian NLU | 66.15 | NLU Average Score (%) | 95.6 |
| EuroEval English NLU - ScaLA EN | 66.22 | Matthews correlation coefficient (%) | 95.2 |
| EuroEval Italian NLU - ScaLA IT | 57.06 | Matthews correlation coefficient (%) | 95 |
| EuroEval Portuguese NLU | 64.05 | NLU Average Score (%) | 94.9 |
| EuroEval French NLU - ScaLA FR | 60.35 | Matthews correlation coefficient (%) | 93.8 |
| EuroEval French NLU | 72.1 | NLU Average Score (%) | 93.6 |
| EuroEval Portuguese NLU - ScaLA PT | 48.4 | Matthews correlation coefficient (%) | 93.4 |
| EuroEval English NLU | 73.27 | NLU Average Score (%) | 93 |
| EuroEval Spanish NLU - ScaLA ES | 43.83 | Matthews correlation coefficient (%) | 92.9 |
| EuroEval Slovene NLU - SSJ500K NER | 72.23 | Micro F1 excluding MISC (%) | 92.6 |
Interactive version: theaggregate.ai/model?slug=nvidia-nemotron-3-nano-30b-a3b-fp8 · How It Works · Data refreshed daily, snapshot 2026-11-08.