NVIDIA Releases Alpamayo 2 Super: A 34B Open-Source VLA Model for Autonomous Driving

drmapavone · x · 2026-08-05

NVIDIA has released Alpamayo 2 Super, an open-source reasoning Vision-Language-Action (VLA) model designed for autonomous vehicles (AV). The 34-billion-parameter model combines the 32B NVIDIA Cosmos 3 Super Reasoner with a 2B diffusion-based Action Expert, and is post-trained using reinforcement learning.

Two key highlights include:

Related event: NVIDIA Releases Open-Source Autonomous Driving VLA Model Alpamayo 2 Super(12 posts)→

Original post →

More from Embodied

Embodied channel →