As of: 2026-11-13 13:49 CET.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/
Canonical page: https://postcutoff.com/m/nemotron-3-ultra/

# NVIDIA Nemotron 3 Ultra (550B-A55B)

NVIDIA, Nemotron 3 family. Status: current. Type: reasoning model. Released: 2026-06-04. Training cutoff: 2025-09.

## How to call it

| Provider | Model id | Endpoint | Docs |
|---|---|---|---|
| NVIDIA API (build.nvidia.com) | `nvidia/nemotron-3-ultra-550b-a55b` | https://integrate.api.nvidia.com/v1/chat/completions | https://build.nvidia.com |
| OpenRouter | `nvidia/nemotron-3-ultra-550b-a55b` |  | https://openrouter.ai/nvidia/nemotron-3-ultra-550b-a55b |
| Hugging Face |  |  | https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 |

## Pricing

- Input: $0.60 per 1M tokens
- Output: $2.40 per 1M tokens
- Unit as published: per 1M tokens (USD) on OpenRouter (262K context there); NVIDIA hosted pricing not verified
- Source: https://openrouter.ai/nvidia/nemotron-3-ultra-550b-a55b
- Last checked: 2026-09-29

## Specs

- Input: text
- Output: text
- Context window: 1,000,000 tokens
- Max output: not published
- Open weights: yes
- Licence: openmdw-1.1

## What it has not seen

- Published training cutoff: 2025-09 (September 2025).
- AI events after it on Post-Cutoff: 1,990 (455 major, 86 historic), as of 2026-11-13.
- Already missing at launch: 168.
- Briefings of the news since September 2025: https://postcutoff.com/briefings/cutoff-2025-09-s.md (short), https://postcutoff.com/briefings/cutoff-2025-09-m.md (medium), https://postcutoff.com/briefings/cutoff-2025-09.md (full).
- Its card, with five things it has never heard of: https://postcutoff.com/you/nemotron-3-ultra/

## What stands out

- Hybrid Mamba-2 / LatentMoE at frontier scale: 550B total / 55B active; interleaved Mamba-2 and LatentMoE layers with select attention, plus multi-token prediction for faster generation. Source: https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16
- NVFP4 pretraining and weights: Pre-trained with an NVFP4 recipe; weights published in both BF16 and NVFP4 under the permissive OpenMDW-1.1 license. Source: https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16
- Reasoning on / off / medium: enable_thinking toggle in the chat template plus a medium-effort mode to cut reasoning tokens. Source: https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16

## Notes

Knowledge cutoff = pre-training data (Sep 2025); post-training data to May 2026. NVFP4 repo nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4.

NVIDIA's largest open reasoning model for agentic workflows and long-context analysis.

```bash
curl https://integrate.api.nvidia.com/v1/chat/completions -H "Authorization: Bearer $NVIDIA_API_KEY" -H "Content-Type: application/json" \
 -d '{"model":"nvidia/nemotron-3-ultra-550b-a55b","messages":[{"role":"user","content":"Hello"}]}'
```

Sources: https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 , https://integrate.api.nvidia.com/v1/models

Model page: https://postcutoff.com/m/nemotron-3-ultra/