Nemotron 3 Nano 30B-A3B

Overview

NVIDIA's 30B-parameter MoE with only 3.5B active per token, delivering 30B-class quality at small-model speeds across chat, code, and reasoning. 128k context.

When to pick this model

• Throughput-sensitive serving where latency matters more than peak quality
• Local inference with partial CPU offload (around 39GB system RAM)
• Long-context reasoning and coding without paying dense-model compute
• Workloads that previously needed a dense 30B but were too slow

VRAM requirements by quantization

VRAM figures include model weights plus a typical 8k KV cache and ~600 MB runtime overhead (Ollama / llama.cpp baseline). Add headroom for higher context lengths.

Strengths

• MoE routing yields 3.5B-class latency with 30B-class capability
• 128k context for large documents and repos
• Strong across chat, code, and reasoning in one checkpoint
• Distillation plus RL alignment from the broader Nemotron family

Limitations

• Needs ~39GB system RAM when partially offloaded to CPU
• NVIDIA Open Model License — review commercial terms
• Gated on Hugging Face

Architecture & training

Architecture: MoE · 30B total / 3.5B active · 128k context

Training: Nemotron 3 family, distillation and RL alignment focused on reasoning, code, and chat.

Quick start

ollama run nemotron-3-nano

Or use the open-source MCP server to query this model from Claude Desktop, Cursor, or any MCP-compatible client.