Nvidia Releases Alpamayo 2 Super: A 34B Open VLA Model for Autonomous Driving

Scobleizer · x · 2026-08-04

Nvidia has unveiled Alpamayo 2 Super, a 34 billion parameter vision-language-action (VLA) reasoning model designed to accelerate autonomous vehicle development.

The model combines the Cosmos 3 Super reasoning model with a 2 billion parameter diffusion-based action expert, post-trained with reinforcement learning. It outputs multiple modalities including future trajectory plans, reasoning traces, grounded scene Q&A, and auto-label generation.

Weights are available on HuggingFace and inference code on GitHub. Distilled models can be deployed commercially without further permission, and outputs carry no license conditions, allowing automakers to easily deploy compact versions on in-car computers.

Related event: NVIDIA Launches Alpamayo 2 Super Open-Source Autonomous Driving Model(7 posts)→

Original post →

More from Embodied

Embodied channel →