Jensen Huang, chief govt officer of Nvidia Corp., speaks subsequent to a BlueField-4 STX Storage tray, from left, Vera CPU compute tray, and Spectrum-X Ethernet Photonics Swap System whereas holding a Spectrum-X Ethernet Photonics chip throughout a keynote handle on the Nvidia GTC convention in San Jose, California, US, on Monday, March 16, 2026.
Bloomberg | Bloomberg | Getty Pictures
In late July, Nvidia CEO Jensen Huang posted on X for the primary time to defend open-source fashions in synthetic intelligence, inserting himself right into a debate that was raging throughout the business.
Lower than three weeks later, Nvidia is releasing Nemotron 3.5 Lightning, which the corporate says is “light-weight” and may run on a single graphics processing unit on a PC. It is Nvidia’s first open-source mannequin since Huang joined most of his tech friends in urging the U.S. authorities to help open fashions whereas “avoiding untimely restrictions” that would push innovation abroad.
The brand new Nemotron providing is free for firms to obtain, use and modify with out getting permission or paying Nvidia.
For Nvidia, open-source AI is a boon for chip gross sales, as a result of the fashions nonetheless must run on GPUs, and the decrease costs can serve to spice up utilization over proprietary fashions from the likes of OpenAI and Anthropic.
“Free AI must be nice for {hardware},” Huang told Axios in an interview final month. “Free AI must be nice for chips.”
Huang jumped headfirst right into a debate that had sprung up in Washington following the announcement of Kimi K3, a mannequin developed by China’s Moonshot AI that narrowed the hole with essentially the most highly effective American fashions. Politicians frightened that Kimi K3 was probably troublesome for nationwide safety, and that it represented mental property theft through a method referred to as distillation, which includes the usage of solutions from a sophisticated AI mannequin’s service to coach a lighter mannequin.
Policymakers questioned whether or not Chinese language mannequin builders may very well be topic to sanctions or different restrictions, much like the foundations positioned on chip gross sales.
Huang posted an open letter, which argued that open-weight fashions permit firms larger management over their future, spur competitors and produce down pricing.
“Open fashions strengthen security and cybersecurity, speed up innovation and diffusion, and allow sovereignty,” Huang wrote in his debut X put up.
Days later, Nvidia launched an AI security consortium with firms together with Microsoft that might concentrate on utilizing open fashions and open-source software program for cybersecurity.
Open-source AI is likely one of the hottest matters in Silicon Valley. On Monday, Meta CEO Mark Zuckerberg printed a prolonged manifesto arguing for open-source AI, as his firm launched a coding mannequin referred to as Muse Spark.
“Our objective must be for American open supply fashions to be the perfect globally,” Zuckerberg wrote.
Nvidia stated firms together with CrowdStrike, CodeRabbit and Harvey have examined and customised its newest mannequin. The chipmaker stated it was developed significantly for brokers, or AI packages that may run autonomously within the background.
Nemotron 3.5 Lightning can even be obtainable on HuggingFace and Nvidia’s web site. Moreover, Nvidia launched software program referred to as NeMo Switchyard that may decide the most cost effective and most applicable AI mannequin for any given activity.
Nvidia representatives stated Monday that the corporate had used distillation to provide Nemotron 3.5 Lightning comparable capabilities to its bigger Nemotron fashions.
