Llama 3.1 Nemotron Nano 4B v1.1: benchmark results
NVIDIA's reasoning and chat model post-trained from Llama-3.1-Minitron-4B-Width-Base, a compressed derivative of Llama 3.1 8B. It supports RAG and tool calling with a publisher-stated 131,072-token context. Provider: NVIDIA. Released 2025-05-20. Access: Open.
Aggregate Rating 1441 ± 136, rank #1020 of 1635 rated models, from 15 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA MATH-500 | 94.67 | Accuracy (%) | 65 |
| AA LiveCodeBench | 49.31 | Pass@1 (%) | 55.6 |
| AA AIME 2025 | 50 | Accuracy (%) | 48 |
| AA Humanity's Last Exam | 5.19 | Accuracy (%) | 29.3 |
| Open Portuguese LLM - OAB Exams | 31.3 | Accuracy (%) | 28 |
| Open Portuguese LLM - FaQuAD NLI | 44.95 | Macro F1 (%) | 27.2 |
| Open Portuguese LLM - ENEM | 30.65 | Accuracy (%) | 24.5 |
| Open Portuguese LLM - BLUEX | 28.37 | Accuracy (%) | 24.1 |
| AA MMLU-Pro | 55.6 | Accuracy (%) | 17.5 |
| Open Portuguese LLM - ASSIN2 RTE | 50.61 | Macro F1 (%) | 15.9 |
| AA GPQA Diamond | 40.81 | Accuracy (%) | 15.5 |
| Tau2-Bench Telecom | 11.7 | Success Rate (%) | 9.5 |
Interactive version: theaggregate.ai/model?slug=llama-3-1-nemotron-nano-4b-v1-1 · How It Works · Data refreshed daily, snapshot 2026-11-05.