NVIDIA Alpamayo 2 Super, the Frontier Open Model for Robotaxis and Autonomous Vehicles, Now Available for Commercial Use
NVIDIA has released Alpamayo 2 Super, an open reasoning model designed for autonomous vehicles and robotaxis, for commercial use. Distributed on Hugging Face under the OpenMDW-1.1 license, the model family enables commercial adaptation, fine-tuning, and deployment. On the LingoQA reasoning benchmark, Alpamayo 2 Super scored 23.2 points higher than GPT-4o using the Lingo-Judge metric.
Key Takeaways
- NVIDIA has released Alpamayo 2 Super, expanding its open reasoning model collection for autonomous vehicles to commercial applications.
Built upon NVIDIA Cosmos 3 Super Reasoner and post-trained using reinforcement learning, the model is hosted on Hugging Face.
- The release applies the Linux Foundation permissive license, OpenMDW-1.1, across the entire Alpamayo model lineup, which includes Alpamayo 1 and Alpamayo 1.5 alongside Alpamayo 2 Super.
This licensing permits commercial redistribution, fine-tuning, and derivative models, giving vehicle manufacturers and suppliers direct ownership of their specialized software and data.
- In performance testing on the LingoQA autonomous driving reasoning benchmark, Alpamayo 2 Super achieved top rank among nearly 40 evaluated systems.
Evaluated via the Lingo-Judge metric, it outperformed GPT-4o by 23.2 points, Gemini 2.5 Pro by 15.1 points, and Qwen2.5-VL 72B by 17.0 points.
- NVIDIA has made its Alpamayo 2 Super autonomous driving reasoning model available for commercial deployment under the OpenMDW-1.1 permissive license.
The OpenMDW-1.1 license allows developers to fine-tune, create derivative models, and commercially redistribute the entire Alpamayo model family.
- On the LingoQA reasoning benchmark, Alpamayo 2 Super outperformed GPT-4o by 23.2 points, Gemini 2.5 Pro by 15.1 points, and Qwen2.5-VL 72B by 17.0 points.
Stats & Key Facts
- #On the LingoQA reasoning benchmark, Alpamayo 2 Super scored 23.2 points higher than GPT-4o using the Lingo-Judge metric.
- #In performance testing on the LingoQA autonomous driving reasoning benchmark, Alpamayo 2 Super achieved top rank among nearly 40 evaluated systems.
- #On the LingoQA reasoning benchmark, Alpamayo 2 Super outperformed GPT-4o by 23.2 points, Gemini 2.5 Pro by 15.1 points, and Qwen2.5-VL 72B by 17.0 points.
NVIDIA has released Alpamayo 2 Super, expanding its open reasoning model collection for autonomous vehicles to commercial applications. Built upon NVIDIA Cosmos 3 Super Reasoner and post-trained using reinforcement learning, the model is hosted on Hugging Face. Developers can use the system in cloud-based workflows to generate synthetic training data, high-quality reasoning traces, and teacher outputs for model distillation into real-time vehicle systems.
The release applies the Linux Foundation permissive license, OpenMDW-1.1, across the entire Alpamayo model lineup, which includes Alpamayo 1 and Alpamayo 1.5 alongside Alpamayo 2 Super. This licensing permits commercial redistribution, fine-tuning, and derivative models, giving vehicle manufacturers and suppliers direct ownership of their specialized software and data. By avoiding full foundation model retraining, organizations can lower frontier-scale development expenses.
In performance testing on the LingoQA autonomous driving reasoning benchmark, Alpamayo 2 Super achieved top rank among nearly 40 evaluated systems. Evaluated via the Lingo-Judge metric, it outperformed GPT-4o by 23.2 points, Gemini 2.5 Pro by 15.1 points, and Qwen2.5-VL 72B by 17.0 points. These results demonstrate how open foundation models can be specialized for domain-specific visual reasoning tasks in safety-critical robotics applications.
NVIDIA has made its Alpamayo 2 Super autonomous driving reasoning model available for commercial deployment under the OpenMDW-1.1 permissive license. The OpenMDW-1.1 license allows developers to fine-tune, create derivative models, and commercially redistribute the entire Alpamayo model family. Built on NVIDIA Cosmos 3 Super Reasoner, the model supports cloud-based development, model distillation, and synthetic data generation for vehicle software.
For more details please read the original article at NVIDIA Blog.
Continue Learning
Comments
Comments appear only after moderation. Your email identifies your submission to the moderator and is never displayed here.
No approved comments yet.