At the Consumer Electronics Show (CES) 2025, NVIDIA officially released its new Cosmos platform, an innovative platform designed to accelerate the development of physical artificial intelligence (AI) systems, especially in the fields of autonomous vehicles and robotics. The Cosmos platform integrates a Generative World Basic Model (WFM), video tagger, security protection mechanism, and an efficient data processing pipeline. These capabilities enable developers to create and create and use more easily without relying on real-world data. Optimize the AI model.
The Cosmos platform will be available in the Hugging Face and NVIDIA NGC catalogs through open model licenses, and will also launch the optimized NVIDIA NIM microservices and provide comprehensive support to enterprises through the NVIDIA AI enterprise software platform. This move will greatly reduce the threshold for developers to use advanced AI technology and promote the birth of more innovative applications.
NVIDIA CEO Jen Hsung said at the show: “Robot technology is about to usher in a turning point like ChatGPT. Like large language models, the basic world model is at the heart of driving the development of robots and autonomous vehicles, but not all developers have it.” abilities and resources to train your own models. We created Cosmos to make the development of physical AI more popular and allow every developer to access general robotics technology. "Huang Renxun's words reveal the core concept of the Cosmos platform- Make AI technology more democratic.
The Cosmos model has the ability to generate physics-based high-definition videos based on text, image and sensor data, making it suitable for a variety of application scenarios such as video search, synthetic data generation and reinforcement learning. Developers can customize models based on specific needs, simulating industrial environments, driving scenarios, and other specific use cases. In addition, NVIDIA has also launched NeMo Curator, an accelerated video processing pipeline that can process 20 million hours of video data in 14 days, and Cosmos Tokeniser, a visual data compression tool that further improves data processing efficiency.
"Data scarcity and variability are key challenges in the successful learning of robotic environments. Cosmos' ability to text, images and video to the world allows us to generate and enhance scenarios for various tasks, thus Training the model without the need for too much expensive real data capture. "This view highlights the unique advantages of the Cosmos platform in solving data problems.
Currently, several major robotics and transportation companies, including Agile Robots, XPENG, Waabi and Uber, have begun to use Cosmos for AI development. "Generating AI will drive future mobility, requiring both rich data and strong computing power," said Dara Khosrowshahi, CEO of Uber. "By working with NVIDIA, we are confident that we can help accelerate safe, scalable autonomous driving solutions," said Dara Khosrowshahi. The cooperation marks the wide recognition of the Cosmos platform in practical applications.
In addition to Cosmos, NVIDIA has also launched the Llama Nemotron large language model and the Cosmos Nemotron visual language model, which is specially developed for enterprises in industries such as healthcare, finance and manufacturing. The launch of these new models further expands NVIDIA's influence in the field of AI and provides enterprises with more customized solutions.
Official blog: https://nvidianews.nvidia.com/news/nvidia-launches-cosmos-world-foundation-model-platform-to-accelerate-physical-ai-development
Key points:
The Cosmos platform aims to accelerate the development of autonomous vehicles and robots and reduce reliance on real data.
Developers can customize models according to their needs and generate video data for multiple application scenarios.
Several robotics and transportation companies have begun to use Cosmos to accelerate the practical application of AI technology.