Nvidia Corp. CEO Jensen speaks while holding a Spectrum-X Ethernet photonics chip next to, from left, a BlueField-4 STX storage tray, a Vera CPU compute tray, and a Spectrum-X Ethernet photonics switch system during the keynote address at the Nvidia GTC conference on Monday, March 16, 2026 in San Jose, California, USA. Mr. Fan.
Bloomberg | Bloomberg | Getty Images
Late July, Nvidia CEO Jensen Huang first posted on X defending the open source model in artificial intelligence, inserting himself into the debate that was brewing across the industry.
Less than three weeks later, Nvidia releases Nemotron 3.5 Lightning. The company says it is “lightweight” and can run on a single graphics processing unit on a PC. This is NVIDIA’s first open source model since Huang joined most of his technology colleagues in pleading with the U.S. government to support an open model while “avoiding premature restrictions” that could push innovation overseas.
New Nemotron products are free for companies to download, use, and modify without obtaining permission or paying Nvidia.
For Nvidia, open source AI will benefit chip sales. This is because the models still need to run on GPUs and the lower price can lead to higher utilization than proprietary models such as OpenAI and Anthropic.
“Free AI should be great for hardware,” Huang said in an interview with Axios last month. “Free AI should be great for chips.”
Huang was the first to jump into the controversy that erupted in Washington after the announcement of Kimi K3, a model developed by China’s Moonshot AI that closes the gap to the most powerful American model. Politicians were concerned that Kimi K3 was potentially troubling to national security and represented theft of intellectual property through a technique known as distillation, which uses answers from advanced AI model services to train lightweight models.
Policymakers questioned whether Chinese model developers could be subject to sanctions and other restrictions similar to those imposed on chip sales.
Huang posted an open letter arguing that the indiscriminate weight model would give companies more control over their futures, increase competition, and lower prices.
“An open model strengthens safety and cybersecurity, accelerates innovation and adoption, and enables sovereignty,” Fan wrote in his debut X post.
A few days later, Nvidia launched an AI Safety Consortium with the following companies: microsoft It will focus on open models and the use of open source software for cybersecurity.
Open source AI is one of the hottest topics in Silicon Valley. on monday, Meta CEO Mark Zuckerberg released a lengthy manifesto advocating for open source AI when his company released a coding model called Muse Spark.
“Our goal should be for America’s open source model to be the best in the world,” Zuckerberg wrote.
According to Nvidia, the following companies cloud strikeCodeRabbit and Harvey tested and customized the latest model. The chipmaker said it was developed specifically for agents, or AI programs that can run autonomously in the background.
Nemotron 3.5 Lightning will also be available on HuggingFace and Nvidia’s website. Additionally, Nvidia has released software called NeMo Switchyard that can determine the cheapest and best AI model for a given task.
Nvidia representatives said Monday that the company used distillation to give the Nemotron 3.5 Lightning features similar to the company’s larger models.
WATCH: Sen. Warner talks about open source AI
