NVIDIA Cosmos 3: Open World Models as Foundation for Physical AI

NVIDIA announced Cosmos 3, a frontier open physical AI foundation omni-model built on a mixture-of-transformers architecture, which combines vision reasoning, world generation, and action prediction, allowing developers to use one model family for scene understanding, synthetic data generation, future state simulation, and building specialized world action models. The family includes Cosmos 3 Super (64B) for high-fidelity world modeling, Cosmos 3 Nano (16B) for efficient reasoning and post-training, and Cosmos 3 Edge (4B) for on-device vision reasoning and robot policy deployment, deployable across NVIDIA RTX GPUs, DGX systems, and Jetson platforms. According to benchmarks, Cosmos 3 ranks No. 1 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, and for robot policy it ranks No. 1 on RoboLab, and Cosmos 3 Super is the highest-ranked open model on VANTAGE-Bench for vision understanding. 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's physical AI stack also includes Isaac GR00T for robotics, Alpamayo for autonomous vehicles, and Metropolis for vision AI. Companies such as Doosan Robotics, LG Electronics, Samsung Electronics, Skild AI, Li Auto, Xiaomi, Afari, Centific, Fogsphere, Linker Vision, Milestone Systems, and Yuan are building on NVIDIA Cosmos for physical AI applications. The NVIDIA Cosmos Coalition, recently expanded to Japan, brings together world model builders, AI developers, and physical AI leaders to contribute models, research, and evaluation methods.