Llama 3.1 Nemotron Nano 8B (v1): benchmark results
NVIDIA's reasoning and chat derivative of Meta Llama 3.1 8B Instruct, post-trained for reasoning, human preferences, RAG and tool calling. The publisher states a 131,072-token context; v1 remains distinct from later Nano releases. Provider: NVIDIA. Released 2025-03-18. Access: Open.
Aggregate Rating 1368 ± 34, rank #1324 of 1635 rated models, from 44 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Swallow - Post-trained English - MATH-500 | 94.6 | Accuracy (%) | 47.8 |
| SpeechMap Compliance | 60.1 | % Requests Completed | 46.3 |
| ZeroEval MATH-500 | 95.4 | MATH-500 Score | 43.8 |
| Open Portuguese LLM - OAB Exams | 39.82 | Accuracy (%) | 42.4 |
| Swallow - Post-trained English - AIME | 57.1 | Accuracy (%) | 41.8 |
| Swallow - Post-trained Japanese - PolyMath High and Top | 28.8 | Accuracy (%) | 41.8 |
| Swallow - Post-trained English - LiveCodeBench | 46.7 | Pass@1 (%) | 40.3 |
| Open Portuguese LLM - ASSIN2 RTE | 85.94 | Macro F1 (%) | 37.2 |
| Open Portuguese LLM - BLUEX | 45.48 | Accuracy (%) | 36.3 |
| Open Portuguese LLM - ENEM | 52.27 | Accuracy (%) | 35.9 |
| Open Portuguese LLM - FaQuAD NLI | 52.81 | Macro F1 (%) | 34.9 |
| Swallow - Post-trained English - Average | 54.3 | Average Score (%) | 28.4 |
Interactive version: theaggregate.ai/model?slug=llama-3-1-nemotron-nano-8b-v1 · How It Works · Data refreshed daily, snapshot 2026-10-15.