NVIDIA
Nemotron models you can run locally
NVIDIA's Nemotron family combines efficient hybrid and mixture-of-experts architectures for local reasoning and agentic workloads.
Available models
Ordered from the lightest model to the largest.
Nemotron Nano 9B v2
9B · Dense
Hybrid Mamba2 architecture for reasoning
128K ctx
Nemotron 3 Nano 30B
30B · 3B active · MoE
MoE with 1M context and 3B active
1024K ctx
Nemotron 3.5 Bolt 74B-A9B
74B · 9B active · MoE
Agentic vision MoE compressed from Nemotron 3.5 Super — 9B active, packed for 24 GB GPUs
256K ctx