Nvidia is opening the gate.
The company released Nemotron 3.5 Lightning, an open-source AI model designed to run on a single GPU inside a laptop or desktop. It is lightweight by design, giving companies a model they can download, use, and modify without asking Nvidia for permission or paying the company.
That combination matters because open-source AI is often discussed as a principle and delivered as a hardware bill. Nemotron 3.5 Lightning addresses both sides of that problem: the model is free to use, and its single-GPU design reduces the hardware required to run it. Nobody has ever confused AI infrastructure with a bargain, so this is a useful change of direction.
A smaller model with larger-model ambitions
Nvidia used distillation to give Nemotron 3.5 Lightning capabilities similar to those found in its larger Nemotron models. The approach lets the new offering remain lightweight without abandoning the capabilities Nvidia associates with its larger models.
The company said CrowdStrike, CodeRabbit, and Harvey have tested and customized the latest model. Those companies now sit among the first named users of Nemotron 3.5 Lightning, showing that Nvidia’s release is aimed at organizations that want to adapt an open model rather than accept a fixed system.
“Free AI should be great for hardware. Free AI should be great for chips,” Nvidia CEO Jensen Huang said. The statement captures the business logic behind the release: Nvidia can support open models while keeping the hardware that runs them at the center of the AI economy.
Nemotron 3.5 Lightning is Nvidia’s first open-source model since Huang joined most of his technology peers in urging the U.S. government to support open models. That makes the release more than another model launch. It is the first concrete open-source model from Nvidia since the company’s leadership made that position clear.
Nvidia’s open-source turn
Huang posted on X in July 2026 to defend open-source models. The post followed his remarks at Nvidia’s GTC conference on Monday, March 16, 2026, and placed Nvidia alongside a broader technology push for open AI development.
Meta CEO Mark Zuckerberg also argued for that direction, saying, “Our goal should be for American open source models to be the best globally.” Microsoft is involved in Nvidia’s AI safety consortium, adding another major company to the surrounding discussion about open models and their development.
The timing gives Nemotron 3.5 Lightning a clear role in Nvidia’s public position. Huang favored open models in July 2026, and Nvidia followed with an open-source model that companies can download, use, and modify without payment or permission. The gap between a policy argument and an available product is where announcements start to earn their keep.
Nvidia’s August 17, 2026, blog post about developing Nemotron 3.5 Lightning NVFP4 adds to that release. The model’s defining promise remains straightforward: similar capabilities to larger Nemotron models, a lightweight design, and operation on a single GPU in a laptop or desktop.
That does not make every AI workload simple, and the facts do not promise that. It does make Nemotron 3.5 Lightning a notable open-source option for companies that want to test and customize Nvidia’s model without paying Nvidia for access. Open AI development has acquired another hardware-friendly recruit.
Based on
