NVIDIA Launches Nemotron 3.5 Lightning and NeMo Switchyard for Agentic AI
NVIDIA expands its Nemotron 3 family with a 30B mixture-of-experts model built for always-on agents, plus an open-source router that sends each request to the best model.
What happened
NVIDIA has expanded its Nemotron 3 family with Nemotron 3.5 Lightning, a 30-billion-parameter mixture-of-experts model it calls the most efficient in its class for long-running agentic AI workloads. It targets specialized tasks inside larger multi-agent systems.
The company also released NeMo Switchyard, an open-source routing library for AI agents. It automatically sends each step of a workflow to the most suitable model and can be tuned for quality, latency or cost — without rewriting existing applications.
Why it matters
Agentic systems increasingly run as "systems of models": a frontier model plans the work while smaller ones handle code review, tool use or security alerts. NVIDIA says Lightning delivers up to 4x faster output and 30 percent faster task completion, while Switchyard keeps frontier-level accuracy at nearly a third of the cost of Opus 4.8 alone.
Nemotron 3.5 Lightning is on Hugging Face, OpenRouter and build.nvidia.com, and runs locally on RTX PCs, DGX Spark and Jetson devices. NeMo Switchyard is on GitHub.