Nvidia launches Alpamayo 2 Super with 32B parameters for autonomous driving

Nvidia's new Alpamayo 2 Super open reasoning model gives autonomous vehicle developers a 32-billion-parameter system that can reason, plan and act across the full driving stack, compressing data annotation cycles from months to days.

"Alpamayo is the moment cars begin to safely reason, not just drive," Jensen Huang, founder and chief executive of Nvidia, said in a statement. "Only Nvidia makes available open models, simulation, real-world data and agent skills so the entire global robotaxi ecosystem can develop level 4 capabilities."

The model scales Nvidia's family from 10 billion to 32 billion parameters and adds full 360-degree surround perception with Meta-Action outputs for high-level driving decisions. Nvidia said reasoning auto-labelling can reduce annotation timelines from months to days, and the model can be distilled into compact forms for on-vehicle hardware deployment. The release sits alongside AlpaGym, an open-source closed-loop reinforcement learning framework that trains models on continuous decision cycles in simulation rather than against static recorded data, exposing compounding errors and edge-case failures that open-loop training misses.

Nvidia also launched OmniDreams, a photorealistic simulation tool for rare and long-tail driving scenarios that standard datasets cannot cover. Neural Reconstruction, powered by Omniverse NuRec, converts real-world fleet footage into 3D scenes adaptable across sensor configurations, reducing the need for repetitive physical data collection. The combined pipeline runs from real-world data capture to in-vehicle deployment.

The autonomous driving push opens a new revenue stream for Nvidia beyond its core data center business, which generated $47.5 billion in the most recent fiscal year. The company's Drive Hyperion platform already counts four new robotaxi partners, and the open-source strategy for Alpamayo 2 Super mirrors the approach that made its CUDA software the standard for AI training — locking developers into Nvidia's ecosystem before competitors such as Qualcomm's Snapdragon Ride and Mobileye's EyeQ can gain traction. Nvidia shares trade at about 35 times forward earnings, and the autonomous driving market represents a total addressable opportunity that Goldman Sachs estimates could reach $1.3 trillion by 2035.

This article is for informational purposes only and does not constitute investment advice.