Nvidia launches a smaller, faster Nemotron model and a router to put it to work

Nvidia introduced Nemotron 3.5 Lightning, a 30B-parameter mixture-of-experts model optimized for speed and customization, alongside NeMo Switchyard, an open-source library for model routing. The new model aims to deliver faster output and better task-specific accuracy through post-training, competing with larger proprietary models in specialized workflows.