nvidia / text-generation updated 3 weeks ago

NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4

:---:--- Total Parameters 550B (55B active) Architecture LatentMoE - Mamba-2 + MoE + Attention hybrid with Multi-Token Prediction (MTP) Context Length Up to 1M tokens Minimum GPU Requirement 4xGB200, 4xB200, 4x GB300, 4x B300, 8xH100 Supported Languages English, French, Spanish, Italian, German, Japanese, Hindi, Korean, Brazilian Portuguese, and Chinese Best For Frontier reasoning, complex agentic workflows, long-con...

Params
302.8 B
Context
262,144
Downloads 30d
345 K
Likes
332
Licence: other Not gated SAFETENSORS 12 languages View on Hugging Face ↗

Download history

daily snapshots · 55 days
▲ 10 K in the last 30 days (2.9%)
441 K209 K
Jul 28Aug 15Sep 2Sep 20

Can you run it?

Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.

FileQuantSizeEst. VRAMVerdict on RTX 4090 · 24 GB
model.safetensors u8 352.3 GB 433.5 GB ❌ Won’t fit
model.safetensors (bf16, full) bf16 + 256K ctx 352.3 GB 1,841.6 GB ❌ Won’t fit
Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.

Run it

copy-paste, exact tags checked against the Hub
~ · curl · api/v1
$ curl -s https://aimodelscomparison.com/api/v1/models/nvidia-nemotron-3-ultra-550b-a55b-nvfp4
{
  "hf_id": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4",
  "params_b": 302.83,
  "context_length": 262144,
  "license": { "id": "other", "commercial": "unknown" },
  "downloads_30d": 345437,
  "vram_estimates": [
    { "quant": "u8", "gb": 433.5 }
  ],
  "updated_at": "2026-08-25T01:00:42Z"
}
est. VRAM —on RTX 4090 · 24 GBJSON API →

Specifications

Architecture
NemotronHForCausalLM
Parameters
302.8 B
Tensor type
U8
Context length
262,144
Vocabulary
131,072
Licence
other
First seen on the Hub
2026-06-03
Training datasets
nvidia/nemotron-post-training-v3, nvidia/nemotron-pre-training-datasets
Added to our catalog
2026-07-28
Compare with any text-generation model