NVIDIA released Nemotron 3.5 Lightning on August 11, the company's first open-source AI model — a 30-billion-parameter mixture-of-experts system designed for autonomous agent workloads. The model is free to download, use, and modify without needing permission or paying NVIDIA.
The release came less than three weeks after NVIDIA CEO Jensen Huang posted on X for the first time to defend open-source AI models, inserting himself into a heated industry debate. "Free AI should be great for hardware. Free AI should be great for chips," Huang told Axios.
The debate was triggered by the release of Kimi K3 from China's Moonshot AI, which narrowed the gap with the most powerful American models and prompted U.S. politicians to question whether Chinese model developers should face sanctions. Huang posted an open letter arguing that open-weight models allow companies greater control over their future, spur competition, and bring down pricing.
Nemotron 3.5 Lightning was developed using distillation from NVIDIA's larger Nemotron models, giving it similar capabilities in a lighter package that can run on a single GPU on a PC. Companies including CrowdStrike, CodeRabbit, and Harvey have tested and customized the model.
Alongside the model, NVIDIA released NeMo Switchyard, an open-source software library that can determine the cheapest and most appropriate AI model for any given task — essentially a traffic router for AI inference.
The release positions NVIDIA at the center of the open-source AI ecosystem it profits from. As the dominant GPU supplier, more open models means more hardware demand. "Our goal should be for American open source models to be the best globally," Meta CEO Mark Zuckerberg wrote the same day, as his company released its own coding model Muse Spark.
