Nemotron 3 Ultra 550B A55B
NVIDIA · nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4
Q4 file 359.2 GB (UD-Q4_K_M) · runs well on a 640 GB multi-GPU server · Weights license: OpenMDW 1.1
Quality
Quality · Overall
LiveBench 2026-06-25 · 0–100 · higher is better · this model outlined, #19 of 23
- Color = publisher
- Memory under review
Source: LiveBench 2026-06-25 · collected 2026-10-18 · open models without a score are not drawn (no score ≠ 0)
- LMArena · text · rating
- 1,426 ± 7Confidence interval 1,419–1,433 · 11,557 votes · rank 122 of the release · Arena name carries (NVFP4), the same NVFP4 weightsLMArena 2026-10-22 · CC BY 4.0
LiveBench categories
This model's mean per category · 0–100
Source: LiveBench 2026-06-25 · collected 2026-10-18 · open models without a score are not drawn (no score ≠ 0)
Reported by the publisher
- Terminal Bench 2.1
- 53.9Agentic
- GDPVal
- 47.9Agentic
- SWE-Bench Verified
- 69.5Agentic
- SWE-Bench Multilingual
- 69.1Agentic
- ProfBench (Search)
- 56.4Agentic
- PinchBench
- 89.8Agentic
- Airline
- 80.0TauBench V3
- Retail
- 88.4TauBench V3
Every reported result
- Telecom
- 93.6TauBench V3
- Banking
- 19.2TauBench V3
- Average
- 70.3TauBench V3
- BrowseComp
- 41.4TauBench V3
- IOI 2025
- 564.7Reasoning and Knowledge
- GPQA (no tools)
- 87.9Reasoning and Knowledge
- SciCode (subtask)
- 43.5Reasoning and Knowledge
- HLE (no tools)
- 26.1Reasoning and Knowledge
- CritPt (no tools)
- 3.4Reasoning and Knowledge
- OmniScience Accuracy
- 24.6Reasoning and Knowledge
- OmniScience Non-Hallucination
- 75.5Reasoning and Knowledge
- IFBench (prompt)
- 82.3Chat & Instruction Following
- AA-LCR
- 65.5Long Context
- RULER 1M
- 94.0Long Context
Across Nemotron 3 sizes
| Benchmark (publisher settings) | Nano 30B A3B | Super 120B A12B | Ultra 550B A55B |
|---|---|---|---|
| IFBench (prompt) | This size's card does not report this result under these settings | 73.30 | 82.3 |
| AA-LCR | This size's card does not report this result under these settings | 58.06 | 65.5 |
Values as printed on the card, with the publisher's own settings (card section: License/Terms of Use › Benchmarks); column “Nemotron 3 Ultra NVFP4”. Not comparable across publishers and not used in any ranking. Model card
Run it
Run it
File UD-Q4_K_M · 359.2 GB · 9 parts · unsloth/NVIDIA-Nemotron-3-Ultra-550B-A55B-GGUF · What is Q4?
llama-server -hf unsloth/NVIDIA-Nemotron-3-Ultra-550B-A55B-GGUF:UD-Q4_K_M- Install llama.cpp
- Run the command above
- Ask at localhost:8080:
Explain what a context window is, in two sentences.
Quality × memory
Quality × memory
LiveBench 2026-06-25 overall vs. Q4 planning memory (log scale) · 19 open models
- This model
- Color = publisher
Memory = the model's own figure: parameters × 0.625 bytes + 20% working room, before any machine's reserve; with a chosen machine, the total there with its reserve (25% of memory, 2 to 8 GB; 2 GB per graphics card). An estimate, not measured VRAM. LiveBench 2026-06-25
Runs on
Within memory, outside the fast part 1
Within the machine's memory, outside its fast part: on a graphics card the layers that do not fit run on the CPU from PC RAM (the listed RAM, or 32 GB assumed); on a Mac, above the macOS default GPU share. Not a speed figure.
Server-class only 1
No single desktop, laptop or GPU profile holds it in fast memory; memory is aggregated across GPUs.
Nemotron 3 Ultra 550B A55B is a text model from NVIDIA with 550B parameters, 55B of them active per token (mixture of experts). It was released on June 4, 2026 and reads up to 262,144 tokens of context. Its Q4 weights plus working room come to about 413 GB, more than the fast memory of any reference Mac or graphics card; of the reference machines only DGX H100 · 8 × 80 GB holds it in fast memory. We collected 23 downloadable GGUF and MLX files from 2 publisher sources; the most downloaded GGUF source is unsloth's (53.0K downloads on the Hub). It scores 67.4 overall in LiveBench 2026-06-25. Its LMArena text rating is 1,426 ± 7 (release 2026-10-22), a separate human-preference scale.
Hardware profiles grouped by the Q4 planning estimate on each machine (dedicated GPUs reserve 2 GB; unified and system memory 25%, 2 to 8 GB): the fast memory holds it, it is within memory but outside the fast part (layers offloaded to PC RAM, or above the macOS GPU share), or the machine is one memory tier short. Each card opens its hardware profile. Not a measured fit or speed.
On the reference machines
| Where Q4 sits | ||||
|---|---|---|---|---|
| Mac mini M2 · 8 GB unifiedLarger than its memoryNo file on the site's quant ladder fits the fast memory at 8,192 tokensNo file in the previous column | ~415 GB | Larger than its memory | No file on the site's quant ladder fits the fast memory at 8,192 tokens | No file in the previous column |
| Mac mini M4 · 16 GB unifiedLarger than its memoryNo file on the site's quant ladder fits the fast memory at 8,192 tokensNo file in the previous column | ~417 GB | Larger than its memory | No file on the site's quant ladder fits the fast memory at 8,192 tokens | No file in the previous column |
| Mac mini M4 · 24 GB unifiedLarger than its memoryNo file on the site's quant ladder fits the fast memory at 8,192 tokensNo file in the previous column | ~419 GB | Larger than its memory | No file on the site's quant ladder fits the fast memory at 8,192 tokens | No file in the previous column |
| Mac mini M4 · 32 GB unifiedLarger than its memoryNo file on the site's quant ladder fits the fast memory at 8,192 tokensNo file in the previous column | ~421 GB | Larger than its memory | No file on the site's quant ladder fits the fast memory at 8,192 tokens | No file in the previous column |
| Mac mini M5 Pro · 64 GB unifiedLarger than its memoryNo file on the site's quant ladder fits the fast memory at 8,192 tokensNo file in the previous column | ~421 GB | Larger than its memory | No file on the site's quant ladder fits the fast memory at 8,192 tokens | No file in the previous column |
| Mac Studio M5 Max · 128 GB unifiedLarger than its memoryNo file on the site's quant ladder fits the fast memory at 8,192 tokensNo file in the previous column | ~421 GB | Larger than its memory | No file on the site's quant ladder fits the fast memory at 8,192 tokens | No file in the previous column |
| Mac Studio · M5 Ultra · 36-core CPU / 80-core GPU · 512 GB unifiedIn memory, outside the fast partUD-IQ4_XS262,144 tokens | ~421 GB | In memory, outside the fast part | UD-IQ4_XS | 262,144 tokens |
| RTX 4060 · 8 GB VRAMLarger than its memoryNo file on the site's quant ladder fits the fast memory at 8,192 tokensNo file in the previous column | ~415 GB | Larger than its memory | No file on the site's quant ladder fits the fast memory at 8,192 tokens | No file in the previous column |
| RTX 3060 · 12 GB VRAMLarger than its memoryNo file on the site's quant ladder fits the fast memory at 8,192 tokensNo file in the previous column | ~415 GB | Larger than its memory | No file on the site's quant ladder fits the fast memory at 8,192 tokens | No file in the previous column |
| RTX 5080 · 16 GB VRAMLarger than its memoryNo file on the site's quant ladder fits the fast memory at 8,192 tokensNo file in the previous column | ~415 GB | Larger than its memory | No file on the site's quant ladder fits the fast memory at 8,192 tokens | No file in the previous column |
| RTX 4090 · 24 GB VRAMLarger than its memoryNo file on the site's quant ladder fits the fast memory at 8,192 tokensNo file in the previous column | ~415 GB | Larger than its memory | No file on the site's quant ladder fits the fast memory at 8,192 tokens | No file in the previous column |
| RTX 5090 · 32 GB VRAMLarger than its memoryNo file on the site's quant ladder fits the fast memory at 8,192 tokensNo file in the previous column | ~415 GB | Larger than its memory | No file on the site's quant ladder fits the fast memory at 8,192 tokens | No file in the previous column |
| DGX H100 · 8 × 80 GBFast memory holds itUD-Q6_K262,144 tokens | ~429 GB | Fast memory holds it | UD-Q6_K | 262,144 tokens |
The file column uses the site's quant ladder (Q8_0 to Q3_K_M, MLX 8 to 4-bit on a Mac): the largest collected file whose size × 1.2 plus the context cache and the machine's reserve fits the fast memory. Planning estimates, not measured runs.
Compare Nemotron 3 Ultra 550B A55B using its publisher card and the LiveBench 2026-06-25 snapshot. This is quality evidence, not local speed.
The Q4 weights plus working room come to about 413 GB from 550B stated parameters, before the machine's reserve (Runs on adds it per machine). It is not a measured minimum; context, batch size, runtime and quantization change actual memory use.
Versions to download
Files you can download for this checkpoint. The section opens on the most downloaded collected GGUF source, unsloth (53.0K Hub downloads), at UD-Q4_K_M. On a Mac it opens on the most downloaded MLX source. Sizes are the Hub's published bytes. What is Q4?
unsloth · 53.0K downloads
unsloth/NVIDIA-Nemotron-3-Ultra-550B-A55B-GGUF · UD-Q4_K_M/NVIDIA-Nemotron-3-Ultra-550B-A55B-UD-Q4_K_M-00001-of-00009.gguf · 359.2 GB
Run this file
llama-server -hf unsloth/NVIDIA-Nemotron-3-Ultra-550B-A55B-GGUF:UD-Q4_K_Mlms get https://huggingface.co/unsloth/NVIDIA-Nemotron-3-Ultra-550B-A55B-GGUF- Ollama: Ollama's documented hf.co form covers single-file GGUFs; this file has 9 parts
- mlx-lm: mlx-lm runs MLX repositories; pick the MLX version on a Mac
Quality scores on this page belong to the original checkpoint; these files are not scored or ranked. File sizes read 2026-10-23 · download counts read 2026-10-17 · commands: Hugging Face Hub docs: Use Ollama with any GGUF model · llama.cpp server README (-hf <user>/<model>[:quant]) · LM Studio docs: lms get with a Hugging Face URL · mlx-lm README (mlx_lm.generate --model)
Fine-tunes, adapters and other versions
- Quantized versions
- 18
- Fine-tunes
- 7
- Adapters · merges
- 14 · 0
Browse them on the Hugging Face model tree · read 2026-10-23
Uncensored and abliterated versions
Some community uploads remove the model's safety refusals. Quality may drop, the original publisher does not support them, claims about them are the uploader's, and the original license still applies. This site never recommends or ranks them.
4 derivative repositories have “uncensored” or “abliterated” in their name (read 2026-10-23).
Measured speed
No published measurement for this checkpoint
Tokens per second are shown only when a source publishes them with hardware, build and setup. They are never estimated.
How to run
The memory needed above is the Q4 load before a machine's reserve. Formula: parameters × 0.625 bytes, plus 20% working allowance; each machine then adds its system reserve: 25% of unified or system memory, 2 to 8 GB, or 2 GB per GPU on dedicated cards. This is neither measured VRAM nor checkpoint download size.
Q4 planning estimate only. Check checkpoint format, runtime support, KV cache, context, batch size and offload before choosing hardware.
Evidence level: Generic estimate only; no execution reproduced by this site.
- Publisher sampling defaults
- temperature 1 · top_p 0.95generation_config.json · read 2026-10-23source
- Chat template markers
- tool calls · thinkingmarkers found in the template, not a test of the feature · read 2026-10-23source
Before downloading
- Check the exact checkpoint and license linked here.
- Confirm a compatible runtime and precision or quantization for your platform.
- Allow for context, batch size and any extra media components beyond the weights.
Technical details
What the source and the catalog store for this checkpoint. Missing fields are left out, never zero.
- Checkpoint / version
- nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4
- Task and modality
- language and reasoning · text
- Count and basis
- 550B · manufacturer specification · 55B active
- Context
- 262,144 tokens published for this model
- Architecture
- MoE
- Released
- nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4 model card (Hugging Face) · read 2026-10-09source
- Runtime and precision
- No local execution verified.
- Languages
- en, fr, es, it, de, pt, ja, ko, hi, ar, zh, heModel card metadata · read 2026-10-25source
- Commercial use
- Allowed OpenMDW 1.1 text (nvidia/NV-Reason-CT LICENSE) · read 2026-10-18license text
Every stored field
- Revision checked
- 02462641f13d3af838b904f48195b9bb8a1e4ebc
- Architecture
- NemotronHForCausalLM
- Library
- transformers
- Hub task tag
- text-generation
- Formats
- safetensors, pt
- Access
- Open download
- Publisher kind
- Official
- Repository created
- 2026-06-03creation is not the model's release
- Repository updated
- 2026-08-24a commit is not a release date
- Hub downloads (30 days) · likes
- 43,690 · 367this repository only · read 2026-10-25
- Read from the Hub page
- transformers class, task tags, license nameHub overview and file listing · read 2026-10-18
- Catalog state
- Hub config or weights found; no inventory benchmark link
550B architectural parameters. The linked checkpoint uses packed NVFP4; its tensor-element count is not its parameter count. The generic Q4 estimate is not a measured requirement for this checkpoint.
Sources
Identity and published specifications come from the checkpoint page. Benchmark scores belong to the named release and do not measure speed on your computer.
- Publisher checkpoint and model card
- Weight license and use terms
- Quality benchmark source · LiveBench 2026-06-25 overall
Catalog reviewed: 2026-09-30. Benchmark release: 2026-06-25; snapshot collected: 2026-10-18.

