NVIDIA Releases Open-Source Alpamayo 2 Edge Self-Driving Model
NVIDIA has released the open-source Alpamayo 2 Edge self-driving model, featuring a specialized optimization by Embedl that runs over 10x faster than PyTorch on edge hardware like AGX Thor. The model takes in data from six cameras and recent vehicle motion to drive autonomous navigation. This release marks a significant milestone for open-source autonomous driving by providing developers with advanced edge-optimized vision-language-action capabilities. The dramatic performance boost on hardware like AGX Thor makes real-time, on-device robotic processing much more viable for production vehicles. Alpamayo 2 Edge is trained using a reinforcement learning loop and processes inputs from six cameras alongside recent motion data. Developers can test and run the optimized model directly on real AGX Thor hardware through Embedl's platform.
## BACKGROUND
NVIDIA Jetson AGX Thor is a powerful GPU-accelerated edge AI platform designed for robotics and autonomous vehicles, utilizing advanced Blackwell architecture and high-performance Arm cores. Embedl provides a deep learning optimization pipeline that compresses and converts standard PyTorch models for efficient hardware deployment.