Nvidia co-founds Open Secure AI Alliance with Microsoft amid open-source push

Nvidia has released its first open-source AI model, marking a concrete step from a company whose CEO Jensen Huang has become one of the most vocal advocates for keeping AI development accessible. The move positions the chipmaker not just as a hardware supplier but as an increasingly central player in the AI model ecosystem itself.

The Alpamayo family and Nvidia’s open AI push

The centerpiece of Nvidia’s open-source offensive is the Alpamayo family of reasoning vision-language-action models. Unveiled at CES 2026 on January 5, these models are purpose-built for Level 4 autonomous driving, the tier where a vehicle handles all driving tasks in specific conditions without human intervention.

Nvidia describes Alpamayo as the world’s first open chain-of-thought reasoning model designed for autonomous vehicle challenges. The models ship with open weights, meaning developers can inspect, fine-tune, and deploy them without licensing hurdles that typically accompany proprietary AI systems.

Alongside the models themselves, Nvidia released AlpaSim, a simulation framework that lets researchers test autonomous driving scenarios in virtual environments before deploying anything on actual roads.

The Alpamayo 2 Super variant received an update around August 4, 2026, specifically enabling commercial use. That distinction matters: an open model you can’t monetize is a research tool, while one you can commercialize is a business enabler.

Huang’s open-source advocacy takes organizational form

On July 24, 2026, Nvidia co-founded the Open Secure AI Alliance alongside Microsoft and other notable technology partners. The coalition’s stated mission is to push back against premature restrictions on open-weight AI models. The alliance argues that open formats actually enhance transparency and safety rather than undermining them.

Huang also signed a public letter advocating for open AI development.

Beyond autonomous driving

Alpamayo isn’t Nvidia’s only open-source bet in 2026. The company also released the Earth-2 family of models targeting weather prediction AI. Nvidia also announced Ising on April 14, 2026, billing it as the world’s first open-source quantum AI model.

Nvidia distributes its open models under commercially friendly licenses including Apache 2.0, MIT, and the Nvidia Open Model License.

Nvidia now ranks as one of the largest contributors on Hugging Face, the platform that has become essentially the GitHub of AI models.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.