NVIDIA puts its Alpamayo driving models under a permissive open commercial license
The company said on August 4 that Alpamayo 2 Super was available for commercial use, and that the OpenMDW license now covered the whole model family rather than research use alone.
NVIDIA said on August 4 that Alpamayo 2 Super, the largest model in its Alpamayo family for automated driving, was available for commercial use on Hugging Face under OpenMDW-1.1, the Linux Foundation license for open AI model distributions. The company said the license covered fine-tuning, derivative models and commercial redistribution.
The licensing change reached beyond the new model. Earlier Alpamayo releases had been introduced for research and development, and NVIDIA said the OpenMDW terms were now being applied across the entire family so that developers could deploy any of the models commercially without seeking further permission.
Alpamayo 2 Super is built on NVIDIA’s Cosmos 3 Super Reasoner and post-trained with reinforcement learning. The company said it offered three times the scale of the 10-billion-parameter Alpamayo 1.5 and Alpamayo 1, reasoned over full-surround camera coverage, and produced five outputs for each driving situation: a planned trajectory, a chain-of-causation trace explaining the decision, a meta-action such as yielding or a lane change, reasoning auto-labels, and visual question answering with two-dimensional grounding.
NVIDIA positioned the model for cloud-based development rather than for the vehicle, describing a workflow in which Alpamayo 2 Super generates reasoning traces, synthetic training data and teacher outputs, and smaller distilled models run in production cars. On the driving reasoning benchmark LingoQA, the company said the model ranked first among nearly 40 evaluated, and that in NVIDIA’s own testing it scored 17.0 points above Qwen2.5-VL 72B, 15.1 above Gemini 2.5 Pro and 23.2 above GPT-4o. It said the Alpamayo family had passed 500,000 downloads on Hugging Face.
No automaker, supplier or production program was named in the announcement.
Source: NVIDIA, August 4, 2026.
Source: NVIDIA