NVIDIA models
Nemotron hybrid reasoning models.
Run on your device
Nemotron 3.5 Lightning
NVIDIA · Aug 2026NVIDIA's newest open reasoning model - a 30B mixture-of-experts with 3B active per token and a 256K context. Strong reasoning, coding, and tool use; runs on a 32 GB machine with the experts in main memory.
from 18.9 GBruns from 32 GB RAM256K contextOpenMDW 1.1
Nemotron 3.5 Bolt
NVIDIA · Oct 2026NVIDIA's edge agent model - a 74B mixture-of-experts with 9B active per token, packed into 22 GB at a native 2-bit quantization, with a 256K context and image understanding. Built for local agents on a 24 GB graphics card, or a 32 GB machine with the experts in main memory.
from 22.0 GBruns from 32 GB RAM256K contextOpenMDW 1.1
Nemotron 3.5 Super VL
NVIDIA · Oct 2026NVIDIA's flagship open agent model - a 120B mixture-of-experts with 12B active per token, a 1M-token context and image understanding. For a workstation with a big graphics card and 96 GB or more of main memory, experts in main memory.
from 71.8 GBruns from 96 GB RAM1M contextOpenMDW 1.1