Skip to content
EN ES
Nvidia

Nemotron 3 Nano

Released 2025-12 · reasoning · 1.0M tokens · 6 benchmarks · Open weight
HuggingFace

Editorial notes

MoE 31.6B / 3.2B activos. Default 256K context, hasta 1M. Supera GPT-OSS-20B y Qwen3-30B-A3B-Thinking-2507.

Spec sheet

Company
Nvidia
Country
US
Type
reasoning
Release
2025-12
Context
1.0M tokens
Params total
31.6B
Params active (MoE)
3.2B
License
nvidia-open-model
Quants
FP16, Q8_0, Q5_K_M, Q4_K_M, Q3_K_M
Pricing (openrouter)
$0.05/$0.2/M
Slug
nemotron-3-nano

Quick install

1 tools
ollama.com 
ollama run nemotron-3:nano

Note: ~18 GB VRAM Q4_K_M

Benchmarks (6)

Reasoning 2

Math 2

Coding 1

Instruction 1

Cite this model
BibTeX · APA

BibTeX 

@misc{frontier-nemotron-3-nano,
  title  = {Nemotron 3 Nano},
  author = {{Nvidia}},
  year   = {2025},
  note   = {Frontier Benchmarks AI atlas. Accessed 2026-05-08},
  url    = {https://frontierbenchmarks.com/models/nemotron-3-nano}
}

APA

Nvidia (2025). Nemotron 3 Nano [Large language model]. Frontier Benchmarks AI. Retrieved 2026-05-08, from https://frontierbenchmarks.com/models/nemotron-3-nano

Citation reflects the atlas page, not the original model paper. For the paper, see the "Resources" section above.

⚔️ Battle vs another model ← All models More from Nvidia Methodology