LLMBOTTLENECK.COM
NVIDIA / nemotron_h

NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16

560.5 billion parameters.

Open in the calculator →

Architecturepublished data
Quick answer
NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 needs about 345.4 GB at Q4_K_M — the format most people download — with 8,192 tokens of context. None of the 18 common devices listed below holds it entirely at this setting. No hardware for it? Rent a GPU that runs it, priced live.

Sized at 8,192 tokens of context with the whole model in device memory, across 18 common devices. Speeds are estimates, not benchmarks.

The config states the stack as a 108-entry layers_block_type and omits num_hidden_layers. The published weight index carries backbone.layers.0 to backbone.layers.107, and the sibling Nemotron 3.5 Lightning checkpoint this catalogue already holds publishes both keys at equal length, so the array is one entry per layer (audit 2026-09-18, A03). Licence declared on the repository: other.

Weight-file size by format

FormatSizeRangeBasis
FP161121.5 GB784.7 GB – 1132.3 GBsize range
Q8_0596.0 GB416.9 GB – 603.5 GBsize range
Q6_K460.3 GB321.9 GB – 467.0 GBsize range
Q5_K_M400.4 GB269.8 GB – 467.0 GBsize range
Q5_0391.2 GB269.8 GB – 467.0 GBsize range
Q4_K_M344.1 GB220.7 GB – 467.0 GBsize range
Q4_0326.2 GB220.7 GB – 467.0 GBsize range
Q3_K_M280.3 GB168.6 GB – 467.0 GBsize range
Q2_K222.0 GB128.7 GB – 467.0 GBsize range

A file size is not the memory a run needs: the KV cache and the runtime reserve come on top, and the calculator adds both.

What the labels mean
published data
Read from a published source — a file's byte count, a model's configuration or a manufacturer's specification — or exact arithmetic on such values. Not a measurement on a machine.
reconstructed size
Weight size reconstructed from the pinned architecture, because no published file exists.
size range
Only a lower and an upper bound are claimed for this weight size.

Hardware ladder

Q4_K_M at 8,192 tokens · 0 of 18 devices hold it entirely
DeviceMemoryVerdictDecodeCalculator
NVIDIA GeForce RTX 5070345.4 GB of 12 GB105 layers on system RAM~0.2 tok/s with offloadOpen →
AMD Radeon™ RX 9070 XT345.4 GB of 16 GB104 layers on system RAM~0.2 tok/s with offloadOpen →
NVIDIA GeForce RTX 4080345.4 GB of 16 GB104 layers on system RAM~0.2 tok/s with offloadOpen →
NVIDIA GeForce RTX 5080345.4 GB of 16 GB104 layers on system RAM~0.2 tok/s with offloadOpen →
NVIDIA GeForce RTX 5070 Ti345.4 GB of 16 GB104 layers on system RAM~0.2 tok/s with offloadOpen →
NVIDIA GeForce RTX 5060 Ti 16GB345.4 GB of 16 GB104 layers on system RAM~0.2 tok/s with offloadOpen →
NVIDIA GeForce RTX 4090345.4 GB of 24 GB101 layers on system RAM~0.2 tok/s with offloadOpen →
AMD Radeon™ RX 7900 XTX345.4 GB of 24 GB101 layers on system RAM~0.2 tok/s with offloadOpen →
NVIDIA GeForce RTX 3090345.4 GB of 24 GB101 layers on system RAM~0.2 tok/s with offloadOpen →
NVIDIA GeForce RTX 5090345.4 GB of 32 GB99 layers on system RAM~0.2 tok/s with offloadOpen →
NVIDIA RTX 6000 Ada Generation345.4 GB of 48 GB94 layers on system RAM~0.2 tok/s with offloadOpen →
Apple M4 Pro345.4 GB of 64 GBdoes not fitnot calibratedOpen →
Apple M5 Pro345.4 GB of 64 GBdoes not fitnot calibratedOpen →
AMD Ryzen AI Max+ 395 with Radeon 8060S345.4 GB of 128 GBdoes not fitnot calibratedOpen →
Apple M3 Max345.4 GB of 128 GBdoes not fitnot calibratedOpen →
Apple M4 Max345.4 GB of 128 GBdoes not fitnot calibratedOpen →
NVIDIA DGX Spark345.4 GB of 128 GBdoes not fitnot calibratedOpen →
Apple M2 Ultra345.4 GB of 192 GBdoes not fitnot calibratedOpen →

Decode is a calibrated estimate from the published calibration; capacity is the manufacturer's published ceiling, not guaranteed free memory. Fitting in memory is not the same as loading: whether the runtime and version you have supports this architecture and format on that machine has not been tested here. Offloaded rows assume there is enough system RAM for the overflow — the calculator checks that against the RAM you declare. Each “Open” link keeps this model, format and context.

Run it in the cloud

Rent a GPU that runs NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16

Every rentable machine that holds the whole model, from Vast.ai and RunPod, with this site's speed estimate for each and the provider's own price, read live. Cheapest first; sort by cost per token or speed instead, or switch the format.

Prices · 18:30 UTC, 22 Sept
6 configurations in stock · 9 out of stock
MachineSpeedPer hourPer M tokensWhereRent
8× L40S384 GBcheapestnot calibrated$3.73—Vast.aiMarketplace · 99.9% reliableRent on Vast.ai
8× RTX PRO 5000384 GBnot calibrated$6.81—Vast.aiMarketplace · 98.3% reliableRent on Vast.ai
2× B300540 GBnot calibrated$15.78—RunPodSecure CloudRent on RunPod
4× H200 NVL564 GBnot calibrated$16.00—Vast.aiMarketplace · 99.1% reliableRent on Vast.ai
4× H200564 GBnot calibrated$18.36—RunPodSecure CloudVast.ai $18.42/h Rent on RunPod
8× H100 PCIe640 GBnot calibrated$19.16—Vast.aiMarketplace · 98.3% reliableRent on Vast.ai
No GPU at all

Use NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 by the token

Nothing to set up and nothing to switch off: you pay only for the tokens you use.

First time renting a GPU? How it works, in four steps
  1. Create an account and add credit. Both providers are prepaid: $10 is enough to try any card on this page for hours. A new RunPod account that arrives through a referral link gets a one-time $5 credit when it first adds $10.
  2. Pick the card and the count shown here, and a template: llama.cpp or vLLM for an OpenAI-compatible API, or one with a web chat if you only want to talk to the model. “How to launch it” below gives the exact image and command.
  3. Wait for the download. The weights are fetched on the machine; tens of gigabytes take a few minutes on a data-centre connection. You pay by the second from the moment it starts.
  4. Stop it when you are done. A running machine bills even when idle. A stopped one costs nothing per hour, though its disk is usually still billed until you delete it.
How to launch it

No Q4_K_M file of NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 is catalogued here, so there is no exact command to vouch for. Start the machine from the llama.cpp server image ghcr.io/ggml-org/llama.cpp:server-cuda and point -hf at a Q4_K_M build from the publisher or a quantizer you trust — search Hugging Face for one. Check its file size against the memory figure on this page before you rent.

Every machine holds the whole model at Q4_K_M and 8,192 tokens of context, no offload. Speeds are this site’s single-stream decode estimates; prices are what each provider’s own API quoted, on-demand, for the whole machine. Vast hosts below 98% measured reliability are left out.

Referral links Vast.ai, RunPod and Novita pay us a share of what you spend if you sign up through these buttons. It costs you nothing, and it never decides an order or a recommendation: both are computed from the live price and the speed, and options that pay us nothing are listed and recommended on the same terms. How we rank

Buying a card instead, or paying by the token? Compare a year of NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 three ways →

Under the hood

Architecture, read from the publisher’s file

The numbers every figure above is computed from, with the file they came from.

Architecture

✓ Architecture read from the published config.json

Retrieved 2026-09-20 at pinned commit 77df655d5e9f; parameter count from the safetensors index at the same revision.

Architecture
nemotron_h
Layers
108
Hidden size
8,192
Attention heads
64
KV heads
2
Head dimension
128
Feed-forward width
5,120
Vocabulary
131,072
Context ceiling
262,144
RoPE theta
10,000
Experts
512
Experts / token
22
Expert width
5,120
Shared experts
1
Shared expert width
10,240

Where the memory goes

tokenembedding× 108 decoder blocksHybrid sequence stack12 attention · 48 Mamba · 48 MLP/MoEfull contextRouted experts22 of 512 per tokenall residentoutputprojectiongrows with contextcapacity ≠ traffic

Grouped-query attention shares each key/value head across 32 query heads, so the KV cache is 3% of what multi-head attention would need at the same context.

Open the exact official config.json ↗

More from NVIDIA

Llama-3.1-Nemotron-Nano-8B-v1Nemotron-3.5-Content-SafetyNemotron-Cascade-2-30B-A3BNVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16NVIDIA-Nemotron-3-Nano-30B-A3B-BF16NVIDIA-Nemotron-3-Nano-4B-BF16NVIDIA-Nemotron-3-Super-120B-A12B-BF16NVIDIA-Nemotron-Nano-9B-v2

Use this answer in your own product

As JSON, from the API
curl -s https://llmbottleneck.com/v1/analyze \
  -H "Authorization: Bearer $LLMB_KEY" -H "content-type: application/json" \
  -d '{"model":"nvidia-nvidia-nemotron-3-ultra-550b-a55b-bf16","quantization":"Q4_K_M","context":8192,"hardware":"nvidia-geforce-rtx-4090"}'

Same engine, same evidence, every field sourced. Free key in one step →

On your page, as a widget
<script src="https://llmbottleneck.com/widget.js"
  data-model="nvidia-nvidia-nemotron-3-ultra-550b-a55b-bf16" data-quantization="Q4_K_M"
  data-hardware="nvidia-geforce-rtx-4090" data-context="8192"></script>

No key needed. Unbranded, with your own buy button, on Pro and Business →

Citing this page

LLM Bottleneck. “NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 VRAM and hardware requirements.” Architecture from nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 at revision 77df655d5e9f, retrieved 2026-09-20. https://llmbottleneck.com/models/nvidia-nvidia-nemotron-3-ultra-550b-a55b-bf16

Every figure above is either the published value or a reconstruction whose measured error is on the accuracy page.

llmbottleneck
catalogue 2026-10-03models 327devices 135