Nvidia has released Nemotron 3.5 Lightning, a customizable open-weight language model designed for local AI agents and high-volume automated workloads. Nemotron 3.5 Lightning uses a hybrid mixture-of-experts architecture combining Mamba-2, expert layers and selected attention layers. Nvidia says Nemotron 3.5 Lightning can run on a single GPU, but hardware requirements vary considerably. This does not imply that the complete model will run on every Nvidia GPU. Nemotron 3.5 Lightning is available through Hugging Face, ModelScope, OpenRouter and Nvidia’s build platform.