NVIDIA unveils open-source AI models to support safe autonomous driving
The Alpamayo family is designed to address so-called “long-tail” driving scenarios, rare and complex situations that remain among the biggest obstacles to large-scale autonomous vehicle deployment
06 January, 2026
TT
16
NVIDIA unveiled a new family of open-source artificial intelligence models, simulation tools and datasets aimed at accelerating the development of safer, reasoning-based autonomous vehicles, as competition intensifies to deploy higher levels of self-driving technology.
Announced at the CES technology show, the Alpamayo family is designed to address so-called “long-tail” driving scenarios, rare and complex situations that remain among the biggest obstacles to large-scale autonomous vehicle deployment.
Autonomous systems have traditionally relied on separate perception and planning models, a structure that can struggle when vehicles encounter unfamiliar conditions.
NVIDIA said Alpamayo introduces reasoning-based vision language action (VLA) models that allow systems to analyse cause and effect step by step, improving decision-making, safety and explainability.
The AI models could help autonomous vehicles handle complex environments
“The ChatGPT moment for physical AI is here, when machines begin to understand, reason and act in the real world,” NVIDIA founder and CEO Jensen Huang said in a statement. He said the technology could help autonomous vehicles handle complex environments and explain their driving decisions, a key factor in building trust and scaling deployment.
The Alpamayo family combines three elements: open AI models, simulation frameworks and large-scale datasets.
Rather than operating directly inside vehicles, the models are designed to act as “teacher” systems, which developers can fine-tune or distil into smaller models suitable for real-world use.
NVIDIA said it is releasing Alpamayo 1, a 10-billion-parameter chain-of-thought reasoning model for autonomous driving research, alongside AlpaSim, an open-source simulation platform for closed-loop testing.
The company is also making available physical AI open datasets comprising more than 1,700 hours of driving data collected across diverse geographies and conditions.
The company said the tools would enable a self-reinforcing development loop, allowing developers to train, test and refine reasoning-based autonomous driving systems more efficiently.
Automotive and mobility companies to explore new tech from NVIDIA
Automotive and mobility companies, including Jaguar Land Rover, Lucid and Uber, as well as research groups such as Berkeley DeepDrive, are exploring the Alpamayo platform, NVIDIA said, as they work toward level 4 autonomy, where vehicles can operate without human intervention in defined conditions.
Against a backdrop of slower progress and rising scrutiny in the autonomous vehicle sector, the company said open development and improved reasoning capabilities could help the industry overcome technical barriers and advance safer deployment at scale.
