NVIDIA Alpamayo
Winning Reason
This open-source platform and product portfolio, specifically engineered for L4 autonomous driving, features the Alpamayo 1.5 VLA inference model, the AlpaSim open simulation framework, and a Physical AI open-source dataset. It empowers manufacturers to build autonomous vehicles with human-like perception, reasoning, and decision-making capabilities while providing the explainability and openness required for safety validation and regulatory collaboration. Utilizing open simulation frameworks and neural reconstruction tools, the platform enables closed-loop testing across diverse traffic scenarios, weather conditions, and edge cases. Seamlessly integrated with the NVIDIA autonomous driving stack, it spans training, simulation, and in-vehicle deployment to foster a robust open-source AI model ecosystem.
Product Feature
Alpamayo is an open ecosystem of models (Alpamayo 1), simulation tools (AlpaSim), and datasets (Physical AI AV Dataset). Alpamayo 1 is a 10-billion-parameter Vision-Language-Action (VLA) model designed to bring human-like thinking to driving. By incorporating step-by-step reasoning, Alpamayo 1 excels at handling unpredictable, rare "long-tail" driving scenarios that traditional AV systems struggle to master. AlpaSim is a fully open-source (Apache 2.0), Python-based, end-to-end simulation framework designed for high-fidelity autonomous vehicle (AV) research and testing. The Physical AI AV Dataset offers developers ~100 terabytes of data representing one of the largest, most geographically diverse collections of multi-sensor data for AV development.