In a significant update, Nvidia has announced the latest artificial intelligence (AI)-powered infrastructure and models “Alpamayo-R1” as it continues building cutting-edge technology for physical AI.
The recently introduced model was announced during the NeurIPS AI conference in San Diego.
According to Nvidia, Alpamayo-R1 is the first model of its kind primarily focused on self-driving systems.
Vision-language models are able to process both photos and text, assisting vehicles to make a better decision by assessing it.
It is built on Nvidia’s Cosmos-Reason model that reasons the decision before responding. Nvidia launched the Cosmos model family in January 2025, with several variants launched in August.
With this addition, Nvidia aims to provide vehicles more “common sense” to tackle complex driving situations.
It is pertinent to mention that the model is currently accessible on GitHub and Hugging Face.
Moreover, the AI-chip manufacturer launched its new guides, inference tools and post-training workflows— called the Cosmos Cookbook — to assist developers customise Cosmos models.
The launch of these updates is part of Nvidia’s efforts to push its boundaries into physical AI, which CEO Jensen Huang and chief scientist Bill Dally described as a significant phase of AI, particularly in robotics.
