NVIDIA Nemotron Nano 9B V2 (Non-reasoning): benchmark results
NVIDIA's hybrid Mamba-2, MLP and attention model with four attention layers and configurable reasoning traces. This catalog row retains the separately evaluated non-reasoning mode. Provider: NVIDIA. Released 2025-08-18. Access: Open.
Aggregate Rating 1321 ± 17, rank #1787 of 2168 rated models, from 195 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA LiveCodeBench | 70.05 | Pass@1 (%) | 80.1 |
| EuroEval Portuguese NLU - HAREM | 48.89 | Micro F1 excluding MISC (%) | 74 |
| EuroEval German NLU - GermEval | 62.81 | Micro F1 excluding MISC (%) | 68.2 |
| EuroEval Estonian NLU - EstNER | 48.15 | Micro F1 excluding MISC (%) | 66.7 |
| EuroEval Italian NLU - MultiNERD IT | 67.43 | Micro F1 excluding MISC (%) | 66.7 |
| EuroEval Dutch NLU - CoNLL NL | 61.51 | Micro F1 excluding MISC (%) | 65.6 |
| EuroEval Spanish NLU - CoNLL ES | 63.43 | Micro F1 excluding MISC (%) | 59.8 |
| EuroEval Danish NLU - Dansk | 54.66 | Micro F1 excluding MISC (%) | 59.5 |
| EuroEval Faroese NLU - FONE | 65.11 | Micro F1 excluding MISC (%) | 58.8 |
| EuroEval English NLU - CoNLL EN | 72.31 | Micro F1 excluding MISC (%) | 58.7 |
| EuroEval Icelandic NLU - MIM-GOLD NER | 50.58 | Micro F1 excluding MISC (%) | 57.8 |
| AA AIME 2025 | 62.33 | Accuracy (%) | 57.7 |
Interactive version: theaggregate.ai/model?slug=nvidia-nemotron-nano-9b-v2-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-20.