Nvidia unveils open AI model for robotaxis at GTC Taipei

NVIDIA Launches Alpamayo 2 Super 32B Reasoning Model for Robotaxis

NVIDIA today unveiled Alpamayo 2 Super at GTC Taipei, a 32-billion-parameter open reasoning vision-language-action model designed to accelerate level 4 robotaxi development. The model triples the parameter count of its predecessor and enables humanlike perception, reasoning, and action across the full autonomous driving stack.barchart

“Alpamayo is the moment cars begin to safely reason, not just drive,” said Jensen Huang, NVIDIA’s founder and CEO. “Only NVIDIA makes available open models, simulation, real-world data and agent skills so the entire global robotaxi ecosystem can develop level 4 capabilities that understand edge cases, explain decisions, earn trust and scale safely to millions of vehicles.”barchart

What Alpamayo 2 Super Delivers

Built on NVIDIA Cosmos world foundation models, Alpamayo 2 Super expands from front-focused cameras to full 360-degree situational awareness and introduces “Meta-Actions” — macro-level driving decisions such as yield, lane change, and stop — alongside trajectory predictions and chain-of-causation traces. The model also introduces reasoning auto-labeling with 2D grounding, compressing annotation cycles from months to days.nvidia

Designed as a teacher model, Alpamayo 2 Super can be distilled into compact models that run on the NVIDIA DRIVE Hyperion platform inside vehicles. Since the Alpamayo family launched, its models have been downloaded close to 400,000 times. The new model is expected to be available this summer on GitHub and Hugging Face.barchart

New Training and Simulation Tools

Alongside the model, NVIDIA introduced AlpaGym, an open-source, high-throughput, closed-loop reinforcement learning framework that runs models through continuous decision-and-observation cycles in simulation — exposing compounding errors and edge-case failures that static datasets miss.automotiveworld

The company also announced OmniDreams, a generative world model for photorealistic closed-loop scenario generation, enabling developers to simulate rare and long-tail driving conditions at scale. New physical AI agent skills powered by NVIDIA Omniverse NuRec allow developers to reconstruct real-world fleet data into photorealistic 3D scenes.barchart

Ecosystem and Partners

The announcements came as part of a broader expansion of the NVIDIA DRIVE Hyperion robotaxi-ready platform ecosystem, bringing together leading global automakers, AV software partners, and ride-hailing mobility providers to build level 4-ready fleets. GTC Taipei, running June 1–4, serves as the stage for these and additional NVIDIA autonomous vehicle announcements.globenewswire