Nvidia unveils Nemotron 3.5 Lightning and NeMo Switchyard to slash enterprise AI costs

Nvidia released Nemotron 3.5 Lightning, a 30-billion-parameter open mixture-of-experts model, and NeMo Switchyard, an open-source Rust library for routing AI workflow steps to the most appropriate model, on August 11, 2026. The model activates only about 3 billion parameters at a time, making it cost-efficient for high-volume agentic tasks, while Switchyard optimizes cost, latency, and capability across models.

Via dwglogo.com Nvidia unveils Nemotron 3.5 Lightning and NeMo Switchyard to slash enterprise AI costs The new 30-billion-parameter model and open-source routing library let companies stop choosing between smart AI and affordable AI Nvidia released Nemotron 3.5 Lightning, a 30-billion-parameter open mixture-of-experts model, alongside NeMo Switchyard, an open-source library that directs each step of an AI workflow to the most appropriate model available. What Nemotron 3.5 Lightning actually does The model uses a mixture-of-experts architecture, which means that while it has 30 billion total parameters, only about 3 billion are active at any given moment. This design makes Nemotron 3.5 Lightning particularly suited for high-volume, specialized agent tasks. The kinds of operations that enterprises run thousands of times per day, like document parsing, data extraction, or customer query classification, don’t need frontier-scale reasoning. Nemotron 3.5 Lightning extends a model family that has been growing steadily since late 2025, when Nvidia rolled out variants including Nano, Super, and Ultra. Each targets a different slice of the performance-cost spectrum, and Lightning slots in as the option optimized for agentic workloads that need to run cheaply at massive scale. NeMo Switchyard: the traffic controller NeMo Switchyard is an open-source Rust library, now available on GitHub, that handles LLM traffic routing and API translations across different models. When an AI agent kicks off a multi-step workflow, Switchyard evaluates each step and routes it to whichever model fits best based on parameters like cost, latency, or capability requirements. A simple classification task might go to Lightning. A complex reasoning step might get routed to a larger frontier model. The release of Nemotron 3.5 Lightning and NeMo Switchyard was made public on August 11, 2026. Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy https://cryptobriefing.com/editorial-policy/ .