Nvidia launches Alpamayo to bring reasoning to autonomous vehicles

At CES 2026, Nvidia launched Alpamayo, a new family of open source AI models, simulation tools, and datasets for training physical robots and vehicles. The technology is designed to help autonomous vehicles reason through complex driving situations. During a press briefing and keynote on Monday, Nvidia CEO Jensen Huang described the launch as “The ChatGPT moment for physical AI is here – when machines begin to understand, reason, and act in the real world.” Huang stated that Alpamayo brings reasoning to autonomous vehicles, allowing them to think through rare scenarios, drive safely in complex environments, and explain their driving decisions.

At the core of this new family is Alpamayo 1, a 10 billion-parameter vision language action (VLA) model. The underlying code for Alpamayo 1 is hosted on Hugging Face, allowing developers to fine-tune the model into smaller, faster versions for vehicle development, train simpler driving systems, or build tools like auto-labeling systems. To support this ecosystem, Nvidia is releasing an open dataset containing more than 1,700 hours of driving data collected across a range of geographies and conditions. Additionally, the company launched AlpaSim, an open source simulation framework hosted on GitHub, which is designed to recreate real-world driving conditions so developers can validate autonomous driving systems.

Alpamayo 1 uses chain-of-thought reasoning—an AI reasoning technique where problems are broken into steps—to help vehicles solve complex edge cases, which are unusual or complex scenarios in autonomous driving. According to Ali Kani, Nvidia’s vice president of automotive, the model operates by breaking down problems into steps, reasoning through every possibility, and then selecting the safest path. Huang noted that the model does not just take sensor inputs to activate the steering wheel, brakes, and acceleration; it also reasons about its actions, explains its decisions, and outlines its trajectory. To further train and test these systems, Kani stated that developers can use Cosmos, Nvidia’s brand of generative world models, to generate synthetic data and combine it with real datasets.

Why it matters

Nvidia’s move to open-source these reasoning models could accelerate the development of autonomous vehicles by allowing developers to tackle complex, unusual driving scenarios that previously required extensive training data.