NVIDIA Corp. (NASDAQ: NVDA) announced the launch of Alpamayo 2 Super, a 32-billion-parameter artificial intelligence model designed for autonomous vehicle development, according to a company press release.

The model represents an expansion from previous 10-billion-parameter versions and incorporates vision-language-action capabilities for level 4 autonomous vehicle development. Alpamayo 2 Super includes full-surround perception across 360 degrees, extending beyond front-facing cameras to include side and rear views.

The company also introduced NVIDIA AlpaGym, a reinforcement learning framework for closed-loop training, and NVIDIA OmniDreams, a generative world model for scenario simulation. These tools are designed to enable developers to test autonomous vehicle models in simulated environments before road deployment.

"Alpamayo is the moment cars begin to safely reason, not just drive," said Jensen Huang, NVIDIA's founder and CEO. The company states the model provides interpretability features for safety validation and regulatory collaboration.

Since its initial launch, the Alpamayo platform has been downloaded approximately 400,000 times, according to NVIDIA. The model includes capabilities for auto-labeling, scene understanding, and trajectory prediction in complex driving scenarios.

Alpamayo 2 Super is designed as a teacher model that can be compressed into smaller versions for deployment on NVIDIA's DRIVE AGX Thor computing platform inside vehicles. The model is expected to become available this summer on GitHub for inference code and on Hugging Face for model weights.

The announcement was made at NVIDIA GTC Taipei. The company recently received recognition from the COMPUTEX Best Choice Awards in the Vehicle Technology and Smart Cockpit category for the Alpamayo platform.