Nvidia just dropped its biggest bet yet on the robotaxi future. Alpamayo 2 Super, a 32-billion-parameter open reasoning model purpose-built for autonomous vehicles, debuted at GTC Taipei on May 31, and it represents a dramatic scale-up from the company’s previous efforts in the space.
The model is designed specifically for level 4 robotaxi development, the tier where vehicles handle all driving tasks without human intervention in defined conditions. It fuses vision, language, and action capabilities into a single system that handles reasoning, planning, and real-time operational decisions across what Nvidia calls the “full autonomous driving stack.”
From 10 billion to 32 billion parameters
To appreciate why this matters, some context helps. The original Alpamayo family, which Nvidia began rolling out earlier in 2026, topped out at around 10 billion parameters. Alpamayo 2 Super more than triples that figure.
Alpamayo 2 Super is what the industry calls a VLA model, short for vision-language-action. Think of it as a system that can see the road, reason about what it sees in something resembling natural language logic, and then translate those conclusions into actual driving maneuvers. Nvidia is making its version freely available as an open model.
