🚗Nvidia Launches Alpamayo 2 Super, A 32B Open Model For Robotaxis
TL;DR
Nvidia debuted Alpamayo 2 Super at GTC Taipei: a 32-billion-parameter open vision-language-action model for level 4 robotaxis. The release bundles AlpaGym for closed-loop RL training and OmniDreams for synthetic scenario generation — a full pipeline from data to deployment.
Nvidia debuted Alpamayo 2 Super at GTC Taipei: a 32-billion-parameter open vision-language-action model for level 4 robotaxis. The release bundles AlpaGym for closed-loop RL training and OmniDreams for synthetic scenario generation — a full pipeline from data to deployment.
Key Points
Alpamayo 2 Super is a 32B-parameter open VLA reasoning model targeting level 4 autonomy
Inference code lands on GitHub and weights on Hugging Face this summer
Companion releases: AlpaGym (closed-loop RL framework) and OmniDreams (generative world model)
JLR, Lucid, Uber and Berkeley DeepDrive are listed early adopters of the broader Alpamayo family
Extends the original Alpamayo line Nvidia introduced at CES 2026 in January
Why It Matters
An open 32B driving brain plus an open simulator narrows the moat Waymo and Tesla have built on internal data — startups can now bootstrap a level 4 stack without renting California miles.
Quick Facts
Frequently Asked Questions
Why does this matter?
An open 32B driving brain plus an open simulator narrows the moat Waymo and Tesla have built on internal data — startups can now bootstrap a level 4 stack without renting California miles.
What happened?
Nvidia debuted Alpamayo 2 Super at GTC Taipei: a 32-billion-parameter open vision-language-action model for level 4 robotaxis. The release bundles AlpaGym for closed-loop RL training and OmniDreams for synthetic scenario generation — a full pipeline from data to deployment.
Comments
Be the first to comment
Enjoyed this article?
Get it daily. 7am. Free. Reads in 5 minutes.
Join 2,645 builders reading daily.