Nemotron-Mini-4B-Instruct: benchmark results
NVIDIA's instruction-tuned derivative of Minitron-4B-Base for English roleplay, retrieval-augmented question answering and function calling. Its parent was pruned and distilled from Nemotron-4 15B; the publisher states a 4,096-token context. Provider: NVIDIA. Released 2024-08-20. Access: Open.
Aggregate Rating 1292 ± 41, rank #8853 of 10916 rated models, from 25 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LatamBoard - Spanish PAWS | 63.2 | Score (%) | 90.9 |
| LatamBoard - Spanish XNLI | 48.23 | Score (%) | 84.8 |
| Open LLM Leaderboard - IFEval | 66.69 | Score | 78.7 |
| LatamBoard - FLORES Bidirectional | 30.02 | Score (%) | 63.6 |
| LatamBoard - Translation Score | 32.62 | Score (%) | 63.6 |
| LatamBoard - Spanish MGSM | 4.8 | Score (%) | 39.4 |
| LatamBoard - Spanish OpenBookQA | 33.2 | Score (%) | 37.9 |
| LatamBoard - BLUEX | 50.35 | Score (%) | 36.4 |
| Open LLM Leaderboard - GPQA | 28.02 | Score | 35.2 |
| LatamBoard | 57.76 | LatamBoard Category Average (percentage) | 33.3 |
| LatamBoard - ENEM Challenge | 60.18 | Score (%) | 33.3 |
| Open LLM Leaderboard - MuSR | 37.67 | Score | 27.7 |
Interactive version: theaggregate.ai/model?slug=nemotron-mini-4b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-08.