Skip to content

Nemotron 3 Ultra 550B A55B on 4× RTX PRO 6000 Blackwell: VRAM, speed and fit

Nemotron 3 Ultra 550B A55B at 4-bit runs on the 4× RTX PRO 6000 Blackwell with 256K of context at about 30 tokens a second.

Text generation · Chat · Ollama / LM Studio

Runs comfortably · on 4× RTX PRO 6000 Blackwell · 4-bit (Q4_K_M) · Ollama / LM Studio · Chat

GPU memory
323 GB
of 384 GB
Context
256K
auto
Speed
30 tok/s
Fast
First reply
0.9 s
2K prompt

Where the memory goes

Model
316 GB
Context
3.4 GB
Overhead
3.0 GB
OS
0.0 GB
Free
61.1 GB
Memory
323 GB of 384 GB
Context
256K tokens · auto
Speed
30 tok/s · fast
First reply
0.9 s for a 2K prompt
Running cost
$33/mo at home · $574/mo on a cloud 8× A40

Validation: Validated

What Nemotron 3 Ultra 550B A55B on 4× RTX PRO 6000 Blackwell was checked against
CheckTargetResultStatus
Fits in GPU memory384 GB usable323 GB neededPass
Headroom5% free61.1 GB free (16%)Pass
Context for chat16K or more256KPass
Writing speed15 tok/s or more30 tok/sPass
First replyunder 3 s, 2K prompt0.9 sPass

Targets: a reply you can read along with, a first reply the use case tolerates, and 5% of memory left free. Calculated from published specs, not benchmarked.

The setup

Model
Nemotron 3 Ultra 550B A55B
Parameters
561B, 55B active
Precision
4-bit (Q4_K_M)
Context
256K, auto
Chats at once
1
Use
Chat
Runs in
Ollama / LM Studio
Hardware
4× RTX PRO 6000 Blackwell
GPU can use
384 GB of 384 GB

Running cost

At home
$33/mo
1520 W at the wall
Renting instead
$574/mo
8× A40 at $4.72/h

4 h a day, power at $0.18 per kWh, cloud at RunPod on-demand prices as of Nov 21, 2026.

How to run it

llama.cpp
llama-server -hf unsloth/NVIDIA-Nemotron-3-Ultra-550B-A55B-GGUF
vLLM
vllm serve nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16
MLX (Mac)
mlx_lm.chat --model mlx-community/Nemotron-3-Ultra-550B-A55B-4bit

Nemotron 3 Ultra 550B A55B on similar machines

Other models that run on 4× RTX PRO 6000 Blackwell

Questions

Can Nemotron 3 Ultra 550B A55B run on 4× RTX PRO 6000 Blackwell?

Nemotron 3 Ultra 550B A55B at 4-bit runs on the 4× RTX PRO 6000 Blackwell with 256K of context at about 30 tokens a second.

Report VR-1KVRKB5, calculated Nov 22, 2026 from published specs, not benchmarked.