Nemotron 3 Ultra — local AI model by NVIDIA
Written by Jakub Rusinowski · Last updated
The top of NVIDIA's Nemotron 3 line: a 550B-parameter hybrid that interleaves Mamba-2 state-space layers with latent mixture-of-experts layers and a minority of attention layers, with 55B active per token. Pre-trained with an NVFP4 recipe and published in BF16 and NVFP4 under OpenMDW-1.1. NVIDIA's own minimum for the BF16 checkpoint is an 8-GPU H200 or B200 server.
Variants
The smallest Nemotron 3 Ultra variant needs about 339 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache.
| Model | VRAM at Q4 | VRAM | Context | Run it |
|---|---|---|---|---|
| Nemotron 3 Ultra 550B → 550B (55B active) | ~339.2 GB | 262,144 | cloud-hosted tag |
Memory is quantized weights plus overhead at Q4_K_M, from the same engine as the GPU & VRAM checker.
How to run Nemotron 3 Ultra locally
Install Ollama, then pull the tag.
No local Ollama tag. Ollama lists nemotron-3-ultra only as a cloud-hosted ':cloud' tag, which runs on Ollama's servers, not your machine.
Pick a size above for its own VRAM figure, speed estimate and install command.
Licence
Commercial use permitted. No usage restrictions beyond attribution.
Applies to: Nemotron 3 Ultra 550BRecommended GPU
The cheapest catalogued GPU that runs Nemotron 3 Ultra locally (min 339 GB VRAM) is the Apple M3 Ultra (512 GB).
Nemotron 3 Ultra — frequently asked questions
How much VRAM does Nemotron 3 Ultra need?
Nemotron 3 Ultra needs about 339 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Nemotron 3 Ultra 550B (339 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run Nemotron 3 Ultra on an RTX 4090 (24 GB)?
Nemotron 3 Ultra's smallest variant needs about 339 GB, which exceeds a single RTX 4090 (24 GB). Use multiple GPUs, a higher-VRAM card, or Apple Silicon with large unified memory.
What quantization should I use for Nemotron 3 Ultra?
Q4_K_M is the best balance of quality and VRAM for Nemotron 3 Ultra in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run Nemotron 3 Ultra with Ollama?
Nemotron 3 Ultra has no local Ollama tag — the published tag is cloud-hosted, so running it sends your prompts to a hosted GPU rather than your own machine.