NVIDIA Nemotron 3 Ultra 550B A55B BF16 — Hardware Requirements & GPU Compatibility
ChatFunctions
NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 is NVIDIA's large hybrid reasoning model, combining Mamba-2, Mixture-of-Experts, and attention layers with multi-token prediction in a single architecture. It has around 560 billion total parameters with about 55 billion active per token, reasoning can be toggled on or off through its chat template, and NVIDIA positions it for frontier reasoning, complex agentic workflows, long-context analysis, and high-stakes retrieval-augmented generation across ten languages. It supports a 262,144-token context window and is released under a custom OpenMDW license. NVIDIA's own stated minimum is 8 GB200/B200-class GPUs or equivalent, so this is squarely server hardware; most people will need a hosted or cloud deployment rather than a local one.
NVIDIA Nemotron 3 Ultra 550B A55B BF16 runs on a 512 GB+ Mac or a multi-GPU server
Best local pick → Mac Studio (M3 Ultra, 512GB) · Q4_K_M · 378.4 GB · ~13 tok/s
Macs from 192 GB (IQ2_XXS · ~25 tok/s)
Score 109.9#26 of 100 open#23 of 96 that fit ≤512 GBOther
Not set up for it?
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
llmrun Composite Score →
109.9
- 90% range 106.8–113.1
- #26 of 100 open models
- #23 of 96 that fit ≤512 GB
Run NVIDIA Nemotron 3 Ultra 550B A55B BF16
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
vllm serve nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 --max-model-len 16384 --tensor-parallel-size 8Size the nodes from the table above
Download: UD-Q4_K_M/NVIDIA-Nemotron-3-Ultra-550B-A55B-UD-Q4_K_M-00001-of-00009.gguf359.23 GB + 8 partsunsloth
Other ways
each opens step-by-stepRun in cloud
Rent 6× H100 SXM · Q4_K_M
- Estimated hourly price
- ~$13.27/hr
- est., whole server
- GPUs
- 6 GPUs
- 480 GB of VRAM
How Much VRAM Does NVIDIA Nemotron 3 Ultra 550B A55B BF16 Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | Runs on | |
|---|---|---|---|---|
| 169.6 GB | 187.6 GB | 2-bit · i-quant extreme compression — Importance-weighted 2-bit, extreme compression — significant quality loss | Runs on B200 192 GB | |
| 185.0 GB | 203 GB | 2-bit · i-quant extra small — Importance-weighted 2-bit, extra small | Runs on Instinct MI325X 256 GB | |
| 262.1 GB | 280.1 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Runs on Instinct MI350X 288 GB | |
| 300.6 GB | 318.7 GB | 3-bit medium — 3-bit medium quantization | Runs on 5× H100 SXM 80 GB | |
| 369.9 GB | 388 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Runs on 6× H100 SXM 80 GB | |
| 439.3 GB | 457.4 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Runs on 7× H100 SXM 80 GB | |
| 508.7 GB | 526.8 GB | 6-bit, very good — 6-bit quantization, very good quality | Runs on 8× H100 SXM 80 GB | |
| 616.6 GB | 634.7 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Runs on 5× H200 SXM 141 GB |
Selected Q4_K_M
4-bit medium quantization378.4GBat 16K context
- Weights
- 368.5 GB
- KV cache
- 9.6 GB
- Overhead
- 0.3 GB
Server-class: ~378.4 GB at 16K context · about 6× H100 SXM 80 GB
Which Hardware Can Run NVIDIA Nemotron 3 Ultra 550B A55B BF16?
Q4_K_M · 369.9 GBNVIDIA Nemotron 3 Ultra 550B A55B BF16 (Q4_K_M) requires 369.9 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 481+ GB is recommended. No consumer GPU has enough memory. 1 device with unified memory can run NVIDIA Nemotron 3 Ultra 550B A55B BF16, including Mac Studio (M3 Ultra, 512GB).
192+ GB
Q4_K_M · best fit IQ2_XXS (178.0 GB)
6 GPUs · 1 Mac class · 2 device families
B200~168Simulate output on B200B300~168Simulate output on B300Instinct MI350X~76Simulate output on Instinct MI350XM3 Ultra256–512 GB~25Simulate output on Mac Studio (M3 Ultra, 256GB)8x H100 SXM5~276Simulate output on NVIDIA DGX H1008x A100 SXM4~234Simulate output on NVIDIA DGX A100 640GB
By VRAM tier · typical GPUs and Macs at Q4_K_M
Consumer GPUs (up to 32 GB) and Macs up to 128 GB: too small for this model (needs ≥ 178 GB)8 GB tier →12 GB tier →16 GB tier →24 GB tier →Macs →
How good is NVIDIA Nemotron 3 Ultra 550B A55B BF16?
The llmrun Score merges 17 public benchmark boards into a single figure, where 100 represents the average for open models.
Score vs VRAM in the ≤512 GB tier
- NVIDIA Nemotron 3 Ultra 550B A55B BF16
- Other models
- Best Score for its VRAM
#23 of 96 that fit ≤512 GB
Each dot is a model at the best quant that fits ≤512 GB. The line joins the models with the best Score for their VRAM; a dot below it is beaten by a model that needs no more memory.
Full ≤512 GB ranking →Benchmarks
The 6 boards that weigh most in the llmrun Score, of 22 with a result for this model; badges show who ran the test.
- GPQA Diamond
Epoch AI · Jun 2026Independent
Independent85.4%
#20/98 open
- LMArena Text
LMArena · Oct 2026Arena
Arena1426.0
#27/187 open
- LiveBench Coding
LiveBench · Jun 2026Independent
Independent70.7
#15/19 open
- LiveBench Reasoning
LiveBench · Jun 2026Independent
Independent74.7
#16/19 open
- AIME 2024/2025
Epoch AI · Jun 2026Independent
Independent86.7%
#18/86 open
- LiveBench Math
LiveBench · Jun 2026Independent
Independent88.7
#6/19 open
In the ≤512 GB class
96 models fit ≤512 GBSame Score, only models that fit ≤512 GB at their best quant: this model's peers by memory need.
- #1
MiMo V2.6 Pro RLXiaomi
133.4
499.3 GB
- #2
MiMo V2.6 Flash RLXiaomi
131.9
186.5 GB
- #3
GLM 5.3Z.ai
130.1
453.8 GB
- #4
DeepSeek V4.1 FlashDeepSeek
129.5
458.4 GB
- #5
DeepSeek V4 Flash 0731DeepSeek
127.6
183.0 GB
- #6
125.3
461.5 GB
Alternatives to NVIDIA Nemotron 3 Ultra 550B A55B BF16
Nemotron family · 42 models →Official models with an llmrun Score, each answering one question; sizes are each model's default quant (usually Q4_K_M) with a 16K-token context.
Best alternative at this size
GLM 5.3
Z.ai
- 20.2 higher Score
- 75.4 GB more VRAM
Score 130.1 · 453.8 GB
Fits in less VRAM
GLM 5.3 Flash
Z.ai
- 14.3 higher Score
- 173.2 GB less VRAM
Score 124.2 · 205.2 GB
Other vendor
DeepSeek V4.1 Flash
DeepSeek
- 19.6 higher Score
- 80.1 GB more VRAM
Score 129.5 · 458.4 GB
Similar models
10| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| GLM 5.2Z.ai | 122.9 | 453.8 GB | —needs > 24 GB | similar score |
| GLM 5.1Z.ai | 117.8 | 454.1 GB | —needs > 24 GB | similar score |
| Qwen3.5 397B A17BAlibaba | 113.0 | 242.6 GB | —needs > 24 GB | similar score |
| MiniMax M3MiniMax | 112.5 | 258 GB | —needs > 24 GB | similar score |
| GLM 5Z.ai | 111.6 | 454.1 GB | —needs > 24 GB | similar score |
| Qwen3.6 27BAlibaba | 111.2 | 17.9 GB | ~38 | similar score |
| DeepSeek V3.2DeepSeek | 110.4 | 412.7 GB | —needs > 24 GB | similar score |
| DeepSeek V4 FlashDeepSeek | 109.3 | 175.1 GB | —needs > 24 GB | similar score |
| GLM 4.7Z.ai | 109.1 | 217.9 GB | —needs > 24 GB | similar score |
| Qwen3 235B A22B Thinking 2507Alibaba | 108.2 | 142.9 GB | —needs > 24 GB | similar score |
Fine-tunes & variants(2)
Specifications
- Publisher
- NVIDIA
- Family
- Nemotron
- Parameters
- 560.5B
- Active parameters
- 57B per token
- Architecture
- NemotronHForCausalLM
- Context Length
- 262,144 tokens
- Vocabulary Size
- 131,072
- Release Date
- 2026-06-03
- License
- Other
Frequently Asked Questions
How much VRAM does NVIDIA Nemotron 3 Ultra 550B A55B BF16 need?
NVIDIA Nemotron 3 Ultra 550B A55B BF16 requires 369.9 GB of VRAM at Q4_K_M, or 1233.2 GB at BF16.
VRAM = Weights + KV Cache + Overhead
Weights = 560.5B × 4.8 bits ÷ 8 = 336.3 GB
KV Cache + Overhead ≈ 33.6 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
VRAM usage by quantization
Can NVIDIA GeForce RTX 5090 run NVIDIA Nemotron 3 Ultra 550B A55B BF16?
No — NVIDIA Nemotron 3 Ultra 550B A55B BF16 requires at least 169.6 GB at IQ2_XXS, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
What's the best quantization for NVIDIA Nemotron 3 Ultra 550B A55B BF16?
For NVIDIA Nemotron 3 Ultra 550B A55B BF16, Q4_K_M (369.9 GB) offers the best balance of quality and VRAM usage. Q5_K_S (423.9 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 169.6 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run NVIDIA Nemotron 3 Ultra 550B A55B BF16 on a Mac?
Yes — Mac Pro M2 Ultra (192 GB) and 3 other Macs can run NVIDIA Nemotron 3 Ultra 550B A55B BF16, the smallest at IQ2_XXS. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 369.9 GB of usable unified memory (RAM minus macOS overhead), from Mac Studio (M3 Ultra, 512GB).
Can I run NVIDIA Nemotron 3 Ultra 550B A55B BF16 locally?
Yes — NVIDIA Nemotron 3 Ultra 550B A55B BF16 can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
What's the download size of NVIDIA Nemotron 3 Ultra 550B A55B BF16?
At Q4_K_M, the download is about 336.31 GB. The full-precision BF16 version is 1121.05 GB. The smallest option (IQ2_XXS) is 154.14 GB.
Which GPUs can run NVIDIA Nemotron 3 Ultra 550B A55B BF16?
No single consumer GPU has enough VRAM to run NVIDIA Nemotron 3 Ultra 550B A55B BF16 at Q4_K_M (369.9 GB). Multi-GPU or professional hardware is required.
Which devices can run NVIDIA Nemotron 3 Ultra 550B A55B BF16?
1 device with unified memory can run NVIDIA Nemotron 3 Ultra 550B A55B BF16 at Q4_K_M (369.9 GB), including Mac Studio (M3 Ultra, 512GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.