Built as the cheap worker beside a larger planner model, so it suits agent builders paying for lots of tool calls.

Kenning's take

What it does

Nemotron 3.5 Lightning is an open 30B mixture-of-experts model with 3B active parameters, built to do the high-volume execution work of AI agents, such as tool calls and checking results, while a larger model plans. It includes built-in speculative decoding for speed and ships with an NVFP4 checkpoint for NVIDIA GPUs. The weights, data and recipes are open, and it can be tried on build.nvidia.com or OpenRouter.

• 30B parameters with 3B active
• Built-in multi-token prediction for faster output
• NVFP4 and BF16 checkpoints
• Fine-tuning with NeMo tools
• Works with NeMo Switchyard model routing

Who it suits. Developers running agents at scale who want a cheap, quick model for routine steps.

Key facts

Category

AI models

Part of

NVIDIA Nemotron 3

Made by

NVIDIA

Based in

United States

Released

11 August 2026

How to get it

weights on Hugging Face and ModelScope; try on build.nvidia.com or through OpenRouter

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.