NVIDIA Cosmos Introduces New Tools for AI Model Development
NVIDIA has expanded its Cosmos platform with new tools to help developers quickly build world models. These models are essential for AI systems that aim to understand and simulate the real world. With the addition of training tools, simulation checkpoints, and blueprints for Cosmos Data Factory, the platform provides a powerful foundation for data, training, and optimization, enabling developers to scale and optimize their models before deploying them in practice.
A crucial aspect of building world models is collecting multimodal and multisensory data. Cosmos Data Factory now offers cloud-native blueprints through platforms like GitHub and Microsoft Azure. These blueprints target various applications, such as generating datasets for autonomous vehicles and enhancing image attributes. Examples of particularly useful applications include creating synthetic videos of rare events and generating defect images for inspection models. Companies like Porsche and ANYbotics are already leveraging these blueprints to accelerate their AI development processes and significantly reduce testing phases.
With the preparation of training data, developers can train their own models or utilize pre-trained models. During the GTC Berlin event, NVIDIA expanded the Cosmos Framework tools for training, optimization, and deployment, accelerating the development of world models. Developers can now customize open models like GR00T and QWEN through fine-tuning and reinforcement learning, as well as optimize and deploy the models with high performance. This makes it a flexible and efficient solution for AI development.
NVIDIA continues to improve its core technology, including new variants of the Cosmos 3 Nano model for simulations. These variants enable developers to test autonomous policies in simulated environments, allowing them to train and evaluate models before deploying them in practice. These innovative techniques apply not only to vehicles but also to robotics and vision AI applications across various industries, including the use of the NVIDIA Metropolis Blueprint for video analysis. The modular setup of the Cosmos platform allows developers to create robust and effective AI solutions.
Read the full article from Nvidia.
Gerelateerde artikelen
European Financial Institutions Shift to Open Models
During GTC Berlin, banks, payment companies, and financial data providers showcase what production-scale intelligence looks like, built on NVIDIA accelerated computing, powered by open-weight models a...
Tech Festival
22 October 2026
NVIDIA DSX: Solution for Energy Bottlenecks
Companies worldwide are utilizing the NVIDIA DSX AI Factory Platform to integrate and optimize energy infrastructure. This platform is designed to enable the grid, facilities, and computational elemen...
Tech Festival
22 October 2026
NVIDIA Transforms European Retail with Open Models
European retailers are employing innovative open models and physical intelligence to enhance their operations. From commercial platforms that reduce processing costs by 90% to fulfillment centers wher...
Tech Festival
22 October 2026