Skip to content

Text model · Nemotron 3

NE Nemotron 3 Ultra 550B-A55B: RAM and VRAM requirements

Nemotron 3 Ultra 550B-A55B needs about 341 GB to run at Q4_K_M (the Q4_K_M GGUF file is ~334.55 GB to download; KV cache and overhead add the rest), or about 550.5 GB at Q8_0. The lightest hardware that runs it is a high-memory machine.

Nemotron 3 family · 550B params (Mixture-of-Experts: activates only 55B of 550B params per token, so generation is faster than the total size suggests) · released Jun 2026.

Shopping for hardware? See what runs Nemotron 3 Ultra 550B-A55B →

License OpenMDW-1.1 · Commercial OK ↓ 451.7K/mo ♥ 302 on HuggingFace
Q4_K_M GGUF
334.55 GB
Q8_0 GGUF
544.13 GB
Memory @ Q4 (4k)
~341 GB
Context
256 k

Q4_K_M and Q8_0 are GGUF download (file) sizes. Memory @ Q4 includes the KV cache and overhead, so it runs larger than the file on disk.

Will it run on your device?

Nemotron 3 Ultra 550B-A55B runs on 0 of 40 tracked devices at Q4_K_M.

0 run well 0 tight fit 40 too small
Fit check Q4_K_M
No, not enough memory
needs 341 GB usable 10.5 GB

Memory breakdown

Weights (Q4_K_M)334.55 GB
+
KV cache (4k)5.6 GB
+
Overhead0.8 GB
=
Total341 GB

How context length changes it

4k context ~341 GB 32k context ~380.4 GB 128k context ~515.4 GB

Longer context grows the KV cache, which for Nemotron 3 Ultra 550B-A55B is sized by its 55B active params, not the full 550B. It needs ~341 GB at 4k and ~515.4 GB at 128k.

Quantization sizes

GGUF quantson disk
Quantization Size on disk
Q2_K 230.3 GB est
Q3_K_M 268.8 GB est
Q4_K_M (default) 334.55 GB
Q5_K_M 391.9 GB est
Q6_K 451 GB est
Q8_0 544.13 GB
FP16 1100 GB est

Lower quant = smaller and faster, slightly lower quality. Q4_K_M is the common default.

Run it

llama.cpp
$ llama-cli -hf unsloth/NVIDIA-Nemotron-3-Ultra-550B-A55B-GGUF:Q4_K_M
LM Studio
$ lms get unsloth/NVIDIA-Nemotron-3-Ultra-550B-A55B-GGUF

Which devices can run Nemotron 3 Ultra 550B-A55B?

Same job, different size

Pick by what fits your memory: step down to free up VRAM, or step up if you have headroom.

Similar models

Head-to-head

FAQ

How much VRAM or RAM does Nemotron 3 Ultra 550B-A55B need?

At Q4_K_M, Nemotron 3 Ultra 550B-A55B needs about 341 GB (weights ~334.55 GB + KV cache + overhead) at a 4k context. At Q8_0 budget ~550.5 GB.

What is the Q4_K_M GGUF file size of Nemotron 3 Ultra 550B-A55B?

The Q4_K_M GGUF file is about 334.55 GB to download, and the Q8_0 GGUF is about 544.13 GB. That is the weights file on disk; to run it you also need room for the KV cache and overhead, so budget ~341 GB of memory at Q4_K_M.

Can Nemotron 3 Ultra 550B-A55B run on a laptop?

Nemotron 3 Ultra 550B-A55B is large; you need a high-memory Mac or multi-GPU setup at Q4_K_M.

Is Nemotron 3 Ultra 550B-A55B cheaper to run because it is a MoE model?

It is faster, not lighter. Nemotron 3 Ultra 550B-A55B activates only 55B of 550B params per token (so it runs quickly), but all experts must stay in memory, so it still needs memory for the full 550B.

Can I use Nemotron 3 Ultra 550B-A55B commercially?

Yes. Nemotron 3 Ultra 550B-A55B is licensed OpenMDW-1.1, which permits commercial use.

Understand the numbers

Short guides to the ideas behind Nemotron 3 Ultra 550B-A55B's memory and quant figures.

NVIDIA's largest Nemotron 3: 550B total / 55B active, Latent-MoE + multi-token-prediction, NVFP4-trained. 256K context by default, up to 1M. Q4 is ~335GB split across 9 GGUF shards (unsloth UD-Q4_K_M); strictly multi-GPU or big-cluster territory, listed for the record like DeepSeek-V4-Pro. Cloud-only on Ollama (no downloadable tag). Linux Foundation's OpenMDW-1.1 license: permissive, free commercial use.

Sources

Last validated 2026-08-03. Memory figures are estimates. See methodology.