Nvidia releases Nemotron 3.5 Lightning open-source AI model

Nvidia has released Nemotron 3.5 Lightning, a lightweight open-source AI model designed to run on a single graphics processing unit in laptops or desktop computers. The release marks the chipmaker's first open model since CEO Jensen Huang publicly entered the debate over open-source artificial intelligence last month.

The model is available for free download, allowing companies to use and modify it without seeking permission or paying licensing fees. Nvidia said the model was developed specifically for AI agents—programs capable of running autonomously in the background—and will be accessible through HuggingFace and Nvidia's website.

Why it matters

Nvidia's embrace of open-source AI reflects a strategic calculation: free models still require GPUs to run, potentially expanding the market for the company's hardware. As Huang told Axios in July, "Free AI should be great for hardware. Free AI should be great for chips." The move also positions Nvidia in a broader industry debate about AI development approaches, with implications for competition, national security, and the balance of power between American and Chinese AI capabilities.

Huang's public stance on open AI

In late July, Huang made his first post on X to defend open-source models, inserting himself into an industry-wide debate that had intensified following the release of Kimi K3, a Chinese model from Moonshot AI that narrowed the performance gap with leading American systems. Policymakers raised concerns about potential intellectual property theft and national security implications.

Huang's open letter argued that open-weight models provide companies with greater control, encourage competition, reduce costs, and strengthen safety and cybersecurity. "Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty," he wrote.

Days after his post, Nvidia launched an AI safety consortium with Microsoft and other companies focused on using open models for cybersecurity applications.

Technical approach and early adopters

Nvidia representatives confirmed the company used distillation—a technique that transfers capabilities from larger models to smaller ones—to give Nemotron 3.5 Lightning performance similar to its larger Nemotron models. The company also released NeMo Switchyard, software that identifies the most cost-effective and appropriate AI model for specific tasks.

Companies including CrowdStrike, CodeRabbit, and Harvey have already tested and customized the model, according to Nvidia.

The release follows Meta CEO Mark Zuckerberg's publication of a lengthy manifesto advocating for open-source AI on the same day, alongside Meta's release of a coding model called Muse Spark. "Our goal should be for American open source models to be the best globally," Zuckerberg wrote.

Details of the Nemotron 3.5 Lightning release were first reported by CNBC.