For teams running long multi-agent workflows that need a million-token context and open weights without per-token lock-in.

Kenning's take

What it does

Nemotron 3 Super is an open reasoning model with 120 billion parameters, 12 billion of them active at a time, and a 1 million token context window. NVIDIA built it to keep long multi-agent workflows on track and to cut the cost of reasoning at every step. The weights are released under a permissive licence, and it is available on build.nvidia.com, OpenRouter, Hugging Face and inside Perplexity.

• 120B parameters with 12B active
• 1 million token context window
• Open weights, data and training recipes
• NVFP4 inference on Blackwell GPUs
• Packaged as an NVIDIA NIM microservice

Who it suits. Companies building agent systems for coding, research or business workflows that want an open model.

Key facts

Category

AI models

Part of

NVIDIA Nemotron 3

Made by

NVIDIA

Based in

United States

Released

11 March 2026

How to get it

open weights on Hugging Face; hosted on build.nvidia.com and OpenRouter; available to Perplexity users

Works on

API, self-hosted

In the UK

Yes

Free plan

free open weights

API price

set by third-party providers

FAQ

Facts from the maker's own pages and the sources listed. Checked 2026-10-11.