NVIDIA is expanding its Nemotron model family with Nemotron 3.5 Lightning, a 30-billion-parameter mixture-of-experts model delivering the highest efficiency for long-running agentic AI workloads. It offers up to 4x faster output speed and 30% faster agentic task completion compared to other models in its class. The open model is customizable for domain-specific tasks using NVIDIA NeMo.

Alongside, NVIDIA released NeMo Switchyard, an open source library for intelligent model routing. It directs each request to the most capable model across a developer's mix of open, proprietary, and NVIDIA models, without requiring application rewrites. Internal benchmarks show it maintains frontier-level accuracy while reducing task completion cost to nearly one-third of using a single frontier model alone.

The combination gives enterprises greater control over AI deployment across PCs, workstations, data centers, and cloud environments. Partners including CrowdStrike, Harvey, CodeRabbit, and others are already customizing Nemotron 3.5 Lightning for specialized tasks like cybersecurity, legal services, and code review. NeMo Switchyard is being integrated into platforms by Boomi, Cadence, Cognition, Kong, LangChain, and more.

Nemotron 3.5 Lightning is available on Hugging Face, Open Router, and as an NVIDIA NIM microservice. NeMo Switchyard is available on GitHub and coming to partner platforms soon.