NVIDIA Unveils New AI Model for Autonomous Vehicles and Robotics

Staff Report: PNN

Semiconductor company NVIDIA on Monday announced a new infrastructure and artificial intelligence model designed to advance physical AI, particularly for robotics and autonomous vehicles. The technology can perceive and monitor real-world environments.

NVIDIA introduced Alpamaio-R-1, an open vision-language model tailored for autonomous driving research. It is the first vision-language model specifically built for autonomous driving, capable of processing both textual and visual information to “see” the surroundings and make decisions.

The model is based on NVIDIA’s Cosmos-Reason model, which performs analysis before making decisions. The Cosmos model family was first released in January 2025, with updates added in August.

According to NVIDIA, Alpamaio-R-1 helps autonomous vehicles achieve four levels of autonomy, enabling full self-governance in specific areas under defined circumstances. The goal is to provide vehicles with human-like reasoning and judgment to make complex and nuanced driving decisions.

The model is currently open on GitHub and Hugging Face, along with Cosmos Cookbook, a comprehensive guide for training, data collection, artificial data generation, and model evaluation.

NVIDIA CEO Jensen Huang has repeatedly stated that the next wave of AI will focus on physical intelligence, where robots and real-world applications play a central role. Chief Scientist Bill Dally emphasized the importance of developing core technology for robotics. This initiative significantly advances the use of AI in real-world applications, particularly in autonomous vehicles and robotics.