NVIDIA Announces New Open AI Models & Tools for Autonomous Driving

Nvidia Advances Physical AI with New Models and Infrastructure

On Monday, Nvidia unveiled new infrastructure solutions and artificial intelligence models. These advancements are geared towards establishing the foundational technology for physical AI, encompassing areas like robotics and self-driving cars.

Introducing Alpamayo-R1: A Vision Language Model for Autonomous Driving

At the NeurIPS AI conference held in San Diego, California, Nvidia introduced Alpamayo-R1. This is an open reasoning vision language model specifically designed for research in autonomous driving.

The company highlights this as the first vision language action model with a primary focus on self-driving capabilities. Such models possess the ability to simultaneously process textual and visual information.

This allows vehicles to interpret their surroundings and formulate decisions based on their perceptions.

Building on the Cosmos Foundation

The Alpamayo-R1 model is built upon Nvidia’s Cosmos-Reason model, a reasoning system that carefully considers options before responding.

The initial Cosmos model family was launched in January 2025, with further models being released in August.

Level 4 Autonomy and Common Sense Reasoning

Nvidia asserts that technologies like Alpamayo-R1 are essential for achieving level 4 autonomous driving. This level signifies full autonomy within a defined area and under specific conditions.

The company’s objective is to equip autonomous vehicles with a form of “common sense,” enabling them to navigate complex driving scenarios with the same nuance as human drivers.

Availability and Developer Resources

The new model is readily accessible on both GitHub and Hugging Face.

Alongside the model, Nvidia has published comprehensive step-by-step guides, inference resources, and post-training workflows on GitHub. This collection, known as the Cosmos Cookbook, aims to assist developers in effectively utilizing and training Cosmos models for their unique applications.

The guide provides coverage of data curation, synthetic data generation, and model evaluation techniques.

The Push Towards Physical AI

These announcements reflect Nvidia’s strong commitment to physical AI as a key growth area for its cutting-edge AI GPUs.

Nvidia’s co-founder and CEO, Jensen Huang, has consistently emphasized that the next significant evolution in AI will be physical AI.

Robotics as a Future Focus

Bill Dally, Nvidia’s chief scientist, reinforced this perspective in a recent conversation with TechCrunch. He highlighted the importance of physical AI, particularly within the field of robotics.

“We aim to be the creators of the ‘brains’ for all robots,” Dally stated. “To achieve this, we must prioritize the development of essential technologies.”

AWS re:Invent 2025 – Live Coverage

The latest innovations in areas like agentic AI, cloud infrastructure, and security are being unveiled at the annual Amazon Web Services event held in Las Vegas.

Event Highlights

AWS’s premier event, re:Invent, is showcasing significant advancements across a broad spectrum of cloud technologies.

Key areas of focus include the development of agentic AI systems and improvements to core cloud infrastructure components.

Sponsored Broadcast

This live stream of AWS re:Invent 2025 is presented as a sponsored broadcast.

• The coverage provides insights into new services and features.
• Attendees and viewers can learn about the future direction of Amazon Web Services.
• The event is a valuable resource for cloud professionals and enthusiasts.

Further details regarding the event’s announcements and keynotes are available through the live broadcast. This video is presented in partnership with AWS.