NVIDIA Releases Its First Autonomous Driving Inference Model, Betting on the Next Generation of AI Brain

On Monday, chip giant NVIDIA announced new infrastructure and artificial intelligence models at the NeurIPS AI conference in San Diego, California. This move aims to accelerate the development of foundational technologies for Physical AI, a field that includes robots and autonomous vehicles capable of perceiving the real world and interacting with it.

The First Autonomous Reasoning Vision-Language Model Debuts

NVIDIA released Alpamayo-R1, an open-source reasoning vision-language model (VLAM) designed for autonomous driving research. The company claims this is the first vision-language action model focused on autonomous driving. Vision-language models can process text and images simultaneously, allowing vehicles to "see" their surroundings and make decisions based on perceptual information.

Alpamayo-R1 is based on NVIDIA's Cosmos-Reason model, a reasoning model that can "think" and make decisions before responding. NVIDIA stated that technologies like Alpamayo-R1 are essential for companies aiming for Level 4 full autonomy and hope that such reasoning models will give autonomous vehicles "common sense," enabling them to handle complex driving decisions more like human drivers.

This new model is now available on GitHub and Hugging Face platforms.

NVIDIA

Cosmos Cookbook: Accelerating Developer Application Deployment

In addition to the new vision model, NVIDIA launched a new set of step-by-step guides, reasoning resources, and post-training workflows on GitHub, collectively known as Cosmos Cookbook. These guides cover data preparation, synthetic data generation, and model evaluation, aiming to help developers better use and train Cosmos models to meet specific application scenario needs.

Focusing on the Next Wave of AI: Physical AI

This release comes as NVIDIA is fully pushing forward in the field of Physical AI, viewing it as a new application direction for its advanced AI GPUs.

Nvidia Releases New AI Model Alpamayo-R1, Advancing Autonomous Driving Research

NVIDIA unveils new AI infrastructure and models at NeurIPS, advancing physical AI for robotics and autonomous vehicles. Highlights include Alpamayo-R1, the first open reasoning vision-language model designed for autonomous driving, enhancing environmental perception by processing text and images.....

NVIDIA CEO Huang Renxun mobilizes employees to use AI tools as much as possible to complete various tasks

NVIDIA CEO Huang Renxun emphasized the application of AI in an internal meeting, requiring employees to fully use AI tools to strengthen the company's leading position in the AI hardware industry and promote internal innovation. After announcing a record-breaking quarterly revenue of 57 billion U.S. dollars, he clearly stated that he aims to expand AI technology to all areas that can be automated.

Google TPU Launch Sparks Price War in AI Computing Market, OpenAI Saves 30% in Costs

Google has shifted from using its own chips internally to selling them externally, launching the TPUv7 'Tie Mu' processor, directly challenging NVIDIA's market position. This move has led to a decline in AI computing prices, changing the previous strategy of internal use only, and actively expanding third-party sales to capture market share.

Million TPU Export Plan Exposed! Google Aims to Sell AI Chips to Meta in 2027, Cloud Revenue May Surge by $13 Billion

Morgan Stanley reports Google plans to sell TPU chips, targeting 1 million units by 2027. 500,000 units expected to add $13B to cloud revenue (11% of total), boosting EPS by 3%. Alphabet shares rose 20% monthly, outperforming Nasdaq. Google shifts from internal use to external sales, with Meta as key client.....