NVIDIA released Cosmos 3, an open physical AI foundation model that enables developers to build specialized systems for

In July, NVIDIA signed an open letter alongside more than 200 companies and organizations affirming that AI leadership will be measured by whether an open ecosystem reaches every sector, not by any single frontier model. Open models, which anyone can download, inspect, modify and run on their own infrastructure, are essential to making this possible, particularly in physical AI where every deployment requires specialization.

Physical AI systems must understand and predict consequences, not just appearances. World models learn how physical environments behave, what may happen next and which actions make sense. They generate physically grounded world and action data, simulate future states and provide foundations that teams can specialize for robots, autonomous vehicles or vision AI systems. Open world models are already generating training data, testing policies and specializing physical AI systems across these domains.

NVIDIA Cosmos 3 brings these capabilities together in an open model family with leading benchmark results and adoption across robotics, autonomous vehicles and vision AI. NVIDIA Omniverse libraries, part of the NVIDIA Agent Toolkit, provide prebuilt capabilities for building simulation-ready worlds that physical AI teams can use to train, test and validate systems before real-world deployment.

The data behind physical AI is difficult and expensive to collect at required scale. World models address this by learning physical relationships from large-scale multimodal scenarios, generating more useful data and providing access to more diverse environments that vary in weather, lighting, objects and trajectories. Since general models have not seen a team's particular robot, sensors or operating environment, closing that gap requires access to model weights, a license permitting adaptation and tools for post-training. NVIDIA Cosmos world foundation models are available under the Linux Foundation's OpenMDW 1.1 license, enabling teams to post-train models on their own data and hardware.

NVIDIA Cosmos 3 is built on a mixture-of-transformers architecture and combines vision reasoning, world generation and action prediction. Developers can use it as a vision language model, as a physics-grounded world simulator that predicts future world states and generates large-scale synthetic data, or as the backbone for world action models, rather than assembling and maintaining separate models for each capability.

The family includes Cosmos 3 Super, a 64-billion-parameter model for high-fidelity world modeling; Cosmos 3 Nano, a 16-billion-parameter model for efficient reasoning and post-training; and Cosmos 3 Edge, a 4-billion-parameter model for on-device vision reasoning and robot policy deployment. Cosmos 3 Edge is lightweight enough to run on edge GPUs, NVIDIA RTX GPUs, NVIDIA DGX systems and NVIDIA Jetson platforms, including Jetson Thor.

Across benchmark evaluations, Cosmos 3 ranks first on Artificial Analysis for open weights text-to-image and image-to-video generation, on PAI-Bench for world generation and in the image-to-video category of Physics-IQ. For robot policy, it ranks first on RoboLab. Cosmos 3 Super is also the highest-ranked open model on VANTAGE-Bench for vision understanding.

The NVIDIA physical AI stack also includes Isaac GR00T for robotics, Alpamayo for autonomous vehicles and Metropolis for vision AI. Developers across industries are building on NVIDIA Cosmos for physical AI applications, including Doosan Robotics, LG Electronics, Samsung Electronics and Skild AI in robotics; Li Auto, Xiaomi and Afari in autonomous vehicles; and Centific, Fogsphere, Linker Vision, Milestone Systems and Yuan for vision AI agents powering industrial AI and smart spaces applications.

The NVIDIA Cosmos Coalition brings together world model builders, AI developers and physical AI leaders to contribute models, research and evaluation methods. NVIDIA recently expanded the coalition to Japan, where robotics and manufacturing leaders intend to join and develop open world models for factories, logistics, agriculture, construction, healthcare and transportation.