NVIDIA Releases Alpamayo 2 Edge End-to-End Driving Model
NVIDIA released an end-to-end autonomous-driving model called Alpamayo 2 Edge. [[fact-klewep4pk]]The model uses an approximately 2.45B-parameter VLM backbone and an approximately 0.2B-parameter action expert. [[fact-v1hupau4h]]It runs on the automotive-grade DRIVE AGX Thor chip. [[fact-yu3v3xivw]]NVIDIA says closed-loop reinforcement learning improved driving performance by 48.9% versus its better checkpoint without that round of reinforcement learning. [[fact-paw3wsq5h]]NVIDIA reported Pixel-to-Trajectory latency of 160–190 milliseconds, depending on the length of the CoC text. [[fact-uyjm1lt19]]That latency excludes the full vehicle response cycle, including sensor capture, localization, controller execution, and safety monitoring. [[fact-fqnmfamfo]]