NVIDIA has expanded its open AI portfolio with Nemotron 3.5 Lightning and NeMo Switchyard, technologies designed to give enterprises greater control over how AI agents run, route workloads, and scale across PCs, edge devices, workstations, data centers, and the cloud.
Nemotron 3.5 Lightning is a 30-billion-parameter mixture-of-experts (MoE) model optimized for high-volume, specialized tasks within multi-agent AI systems. NVIDIA says the model can deliver up to 4x faster output speeds and up to 30% faster agentic task completion than other models in its class.
Because the model is open and customizable, organizations can post-train it using their own data, tools, and workflows through NVIDIA NeMo. This allows enterprises to tailor AI agents for specialized applications such as code review, cybersecurity monitoring, tool use, and customer or billing support.
Nemotron 3.5 Lightning can also run locally on systems including NVIDIA RTX PCs, DGX Spark, DGX Station, and Jetson, as well as across RTX PRO workstations, data centers, and cloud environments. This deployment flexibility can help organizations maintain greater control over data privacy and infrastructure.
Alongside the model, NVIDIA introduced NeMo Switchyard, an open-source routing library designed to automatically direct AI-agent requests to the most suitable model for each task. Developers can configure routing based on priorities such as accuracy, latency, and cost while using a mix of open, proprietary, and NVIDIA models.
NVIDIA says internal benchmarks showed Switchyard maintaining frontier-level accuracy while reducing task-completion costs to nearly one-third of using a single frontier model. Partners including Boomi, LangChain, Ramp, Cognition, and Classmethod have also reported cost or latency improvements in their testing and deployments.
Together, Nemotron 3.5 Lightning and NeMo Switchyard reflect a shift toward systems of AI models, where different models handle specialized tasks rather than relying on a single model for an entire workflow. The approach gives enterprises more flexibility to optimize AI performance, costs, privacy, and deployment as agentic AI moves into production.
Nemotron 3.5 Lightning is available through Hugging Face, ModelScope, OpenRouter, and NVIDIA’s developer platform, while NeMo Switchyard is available as an open-source project on GitHub.
