{"id":109305,"date":"2026-08-05T20:23:58","date_gmt":"2026-08-05T20:23:58","guid":{"rendered":"https:\/\/youzum.net\/nvidia-releases-alpamayo-2-super-a-34b-open-vision-language-action-model-for-robotaxis-and-autonomous-driving-under-openmdw-1-1\/"},"modified":"2026-08-05T20:23:58","modified_gmt":"2026-08-05T20:23:58","slug":"nvidia-releases-alpamayo-2-super-a-34b-open-vision-language-action-model-for-robotaxis-and-autonomous-driving-under-openmdw-1-1","status":"publish","type":"post","link":"https:\/\/youzum.net\/th\/nvidia-releases-alpamayo-2-super-a-34b-open-vision-language-action-model-for-robotaxis-and-autonomous-driving-under-openmdw-1-1\/","title":{"rendered":"NVIDIA Releases Alpamayo 2 Super: A 34B Open Vision-Language-Action Model for Robotaxis and Autonomous Driving Under OpenMDW-1.1"},"content":{"rendered":"

NVIDIA has released Alpamayo 2 Super<\/a><\/strong>, a 34B-parameter vision-language-action (VLA) mode<\/a>l for autonomous driving, under an open commercial license. The stated design target is the long-tail events: rare, multi-agent situations that conventional detection-and-prediction stacks handle poorly. The model pairs a 32B VLM backbone, built on NVIDIA Cosmos 3 Super Reasoner and post-trained with reinforcement learning, with a 2.3B diffusion-based action decoder. From one pass over full-surround camera video it emits a planned trajectory, a causal explanation of that trajectory, and a meta-action. <\/p>\n

Is it deployable<\/strong><\/h2>\n

Yes, and for commercial use from day one. The weights are released under OpenMDW-1.1, the Linux Foundation\u2019s permissive license for open model distributions; source code is Apache 2.0. The license covers fine-tuning, derivative models and commercial redistribution. NVIDIA is applying OpenMDW across the entire Alpamayo family, so earlier releases introduced for R&D are now deployable commercially without additional permission.<\/p>\n

Inputs, outputs and training data<\/strong><\/h2>\n

Inputs are multi-camera RGB video, text, and egomotion history with timestamps. The validated public notebook profiles use six cameras and four historical frames per camera. Egomotion is 3D translation plus a 3\u00d73 rotation matrix, multi-timestep.<\/p>\n

The trajectory API returns 64 waypoints spanning 0.1 to 6.4 seconds at 0.1-second intervals. Each waypoint carries ego-frame XYZ and a 3\u00d73 rotation matrix.<\/p>\n

Training data is roughly 115,000 hours of multi-camera driving video with egomotion and trajectory annotations. It includes about 3,700,000 Chain-of-Causation (CoC) traces \u2014 structured, causally linked explanations of driving decisions. Image training data exceeds one billion images.<\/p>\n

Benchmarks<\/strong><\/h2>\n

On LingoQA<\/a>, Alpamayo 2 Super records a Lingo-Judge score of 79.2 and ranks first among nearly 40 models evaluated. In NVIDIA\u2019s testing it beat Qwen2.5-VL 72B by 17.0 points, Gemini 2.5 Pro by 15.1, and GPT-4o by 23.2.<\/p>\n

Two more numbers matter for planning work. Closed-loop evaluation with AlpaSim<\/a> on 910 scenarios from the PhysicalAI-AV-NuRec<\/a> dataset gives an AlpaSim score of 1.50 \u00b1 0.13. Open-loop evaluation on 937 challenging samples from the PhysicalAI-AV<\/a> dataset gives minADE\u2086 at 6.4s of 0.911m.<\/p>\n

Five outputs from one model<\/strong><\/h2>\n

For each driving situation, the model produces a trajectory, a CoC trace explaining the decision, a meta-action such as yield or lane change, reasoning auto-labels, and visual question answering with 2D grounding.<\/p>\n

That combination is what makes the release interesting operationally. Developers can tie what the model observed to the action it chose. CoC traces integrate with NVIDIA Halos<\/a> safety-validation workflows and support AI safety aligned with ISO\/PAS 8800.<\/p>\n

Used as an autolabeler on proprietary fleet data, NVIDIA says the model compresses annotation cycles from months to days.<\/p>\n

Interactive explainer<\/strong><\/h2>\n
\n<\/div>\n

\n

Key Takeaways<\/strong><\/h2>\n