Nvidia launches Alpamayo, open AI models that allow autonomous vehicles to ‘think like a human’

At CES 2026, Nvidia launched Alpamayo, a caller household of open-source AI models, simulation tools, and datasets for grooming carnal robots and vehicles that are designed to assistance usher autonomous vehicles crushed done analyzable driving situations.

“The ChatGPT infinitesimal for carnal AI is present – erstwhile machines statesman to understand, reason, and enactment successful the existent world,” Nvidia CEO Jensen Huang said successful a statement. “Alpamayo brings reasoning to autonomous vehicles, allowing them to deliberation done uncommon scenarios, thrust safely successful analyzable environments, and explicate their driving decisions.” 

At the halfway of Nvidia’s caller household is Alpamayo 1, a 10-billion-parameter chain-of-thought, reason-based imaginativeness connection enactment (VLA) exemplary that allows an AV to deliberation much similar a quality truthful it tin lick analyzable borderline cases – similar however to navigate a postulation airy outage astatine a engaged intersection – without erstwhile experience. 

“It does this by breaking down problems into steps, reasoning done each possibility, and past selecting the safest path,” Ali Kani, Nvidia’s vice president of automotive, said Monday during a property briefing. 

Alpamayo 1’s underlying codification is disposable connected Hugging Face. Developers tin fine-tune Alpamayo into smaller, faster versions for conveyance development, usage it to bid simpler driving systems, oregon physique tools connected apical of it similar auto-labeling systems that automatically tag video information oregon evaluators that cheque if a car made a astute decision. 

“They tin besides usage Cosmos to make synthetic information and past bid and trial their Alpamayo-based AV exertion connected the operation of the existent and synthetic dataset,” Kani said. Cosmos is Nvidia’s marque of generative satellite models, AI systems that make a practice of a carnal situation truthful they tin marque predictions and instrumentality actions. 

As portion of the Alpamayo rollout, Nvidia is besides releasing an unfastened dataset with much than 1,700 hours of driving information collected crossed a scope of geographies and conditions, covering uncommon and analyzable real-world scenarios. The institution is additionally launching AlpaSim, an open-source simulation model for validating autonomous driving systems. Available connected GitHub, AlpaSim is designed to recreate real-world driving conditions, from sensors to traffic, truthful developers tin safely trial systems astatine scale.