NVIDIA Alpamayo 2 Super

NVIDIA Alpamayo 2 Super is an open 34-billion-parameter reasoning vision-language-action (VLA) model for autonomous driving. It pairs the 32B Cosmos 3 Super Reasoner with a roughly 2B diffusion-based Action Expert and is post-trained with reinforcement learning. From a single pass over 360-degree video from up to seven cameras plus egomotion history, it emits a planned trajectory, a Chain-of-Causation trace explaining the decision, a meta-action such as yield, lane change or stop, reasoning auto-labels, and visual question answering with 2D grounding. It ranks first on the LingoQA driving-reasoning benchmark among nearly 40 models with a Lingo-Judge score of 79.2, beating Gemini 2.5 Pro by 15.1 points and GPT-4o by 23.2 points. Weights are on Hugging Face under the Linux Foundation's OpenMDW-1.1 licence with source code under Apache 2.0, permitting fine-tuning, derivative models and commercial redistribution. NVIDIA positions it as a cloud-scale teacher model to be distilled for in-vehicle inference rather than run directly in the car.

Benchmark Performance