NVIDIA Releases Alpamayo 2 Super: A 34B Open Vision-Language-Action Model for Robotaxis and Autonomous Driving Under OpenMDW-1.1

NVIDIA has introduced Alpamayo 2 Super, a 34-billion-parameter vision-language-action (VLA) model designed specifically for autonomous driving. Released under an open commercial license, the model targets the long tail of rare, multi-agent traffic scenarios that conventional detection-and-prediction systems often struggle to handle. Alpamayo 2 Super pairs a 32B vision-language backbone, built on NVIDIA Cosmos 3 Super Reasoner and post-trained with reinforcement learning, with a 2.3B diffusion-based action decoder. From a single pass over full-surround camera video, the model outputs a planned trajectory, a causal explanation of that trajectory, and a meta-action. ## Is It Deployable? Yes, and it is available for commercial use from day one. The model weights are distributed under OpenMDW-1.1, the Linux Foundation's permissive license for open model distributions, while the source code is Apache 2.0. This license permits fine-tuning, derivative models, and commercial redistribution. NVIDIA is applying OpenMDW across the entire Alpamayo family, meaning earlier research-oriented releases are now commercially deployable without additional permissions. ## Inputs, Outputs, and Training Data The model accepts multi-camera RGB video, text, and ego-motion history with timestamps as inputs. Validated public notebook profiles use six cameras and four historical frames per camera. Ego-motion is represented by 3D translation and a 3×3 rotation matrix at multiple timesteps. The model's output includes a trajectory plan, an explanation, and a meta-action, all generated end-to-end.

via MarkTechPost