**Physical AI** # NVIDIA Cosmos Develop physical AI faster with leading world foundation models and open data processing, training, and evaluation frameworks. [Download Models](https://huggingface.co/collections/nvidia/cosmos3) [Try Now](https://build.nvidia.com/models?q=cosmos)   |   [Video](https://www.youtube.com/watch?v=q7Hj3J9SOXw)   |   [Blog](https://blogs.nvidia.com/blog/cosmos-3-physical-ai-open-world-foundation-model) Cosmos 3 ## The Open Physical AI Foundation Model The first [omni-model](https://www.nvidia.com/en-us/glossary/omni-model.md) with native reasoning, world and action generation. Built on [Mixture-of-Transformers](https://www.nvidia.com/en-us/glossary/mixture-of-transformers.md). [ ](https://images.nvidia.com/aem-dam/Solutions/ai/cosmos/cosmos-power-vision-ai-reasoning-video.mp4) ### Power Vision AI Reasoning Use as a vision language model (VLM) to reason over objects, interactions, and intent across complex real-world scenarios. For real-time alerts and dense captioning across quality inspection, public safety, traffic monitoring, logistics, and autonomous driving. [ ](https://images.nvidia.com/aem-dam/Solutions/ai/cosmos/cosmos-build-policy-models-video.mp4) ### Build Policy Models Accelerate robot policy learning with NVIDIA Cosmos™ 3 as the backbone for World Action Models (WAMs). Post-train the generalized world foundation model on specialized camera and embodiment data. The policy model adapts pre-learned actions to specific tasks, domains, and behaviors at scale. [ ](https://images.nvidia.com/aem-dam/Solutions/ai/cosmos/cosmos-simulate-worlds-video.mp4) ### Simulate Worlds Run as a controllable, physics-grounded world simulator to predict multiple approaches, evaluate outcomes in a closed loop, and converge on the right behavior. Scale the loop across environments, tasks, and conditions to continuously improve without real-world risk. [ ](https://images.nvidia.com/aem-dam/Solutions/ai/cosmos/cosmos-scale-synthetic-video-data-video.mp4) ### Scale Synthetic Video Data Generate infinite plausible futures from text, image, video, ambient sound and action input. Use video generation as imagination to train physical AI without being constrained by what's been physically captured. [Learn More](https://research.nvidia.com/labs/cosmos-lab/cosmos3/technical-report.pdf) Video ## Introducing Cosmos ![YouTube Video](https://img.youtube.com/vi_webp/1QPh70Es_oU/maxresdefault.webp) Consent for Optional Cookies (googleCookiePolicyLink)YouTube sets performance, advertising, and other optional cookies(/googleCookiePolicyLink) when you watch embedded videos. To watch this video, you need to turn on optional cookies for the site. By clicking “Accept and Play Video,” you will automatically turn on advertising and other optional cookies for the site and accept our (nvidiaTermsOfServiceLink)Terms of Service(/nvidiaTermsOfServiceLink) (which contains important waivers). Please see our (nvidiaPrivacyPolicyLink)Privacy Policy(/nvidiaPrivacyPolicyLink) and (nvidiaCookiePolicyLink)Cookie Policy(/nvidiaCookiePolicyLink) for more information. Cancel Accept and Play Video Alternatively, you can (youtubeLink)watch this video on YouTube(/youtubeLink). Meet NVIDIA Cosmos—the frontier of physical AI. Built on a new mixture-of-transformers architecture, this open omnimodel brings together pixels, actions, sound, and language to reason, plan, and generate what comes next. [Watch Video](https://youtu.be/1QPh70Es_oU) Starting Options ## Get Started With NVIDIA Cosmos #### 1 Ready to build? Access open models and code directly. [Download Models](https://huggingface.co/collections/nvidia/cosmos3) [Customize Models](https://github.com/nvidia/Cosmos) #### 2 Not ready to build yet? Try Cosmos models in our hosted catalog. [Try Now](https://build.nvidia.com/search?q=cosmos) #### 3 Need help? Start quickly with our hands-on model recipes. [Browse Cookbook](https://nvidia-cosmos.github.io/cosmos-cookbook/) ## Develop With Cosmos Build on the same technology powering Cosmos 3. Open frameworks and skills so developers worldwide can customize, extend, and contribute to physical AI. ### Data Curation Quickly filter, annotate, and deduplicate large amounts of sensor data with Cosmos Curator. [Download Cosmos Curator on GitHub](https://github.com/nvidia-cosmos/cosmos-curate) Review and score generative video outputs at scale using Cosmos Evaluator. [Download Cosmos Evaluator on GitHub](https://github.com/nvidia-cosmos/cosmos-evaluator) ### Training and Acceleration Quickly build, post-train or deploy world models using open post-training, evaluation, optimization frameworks, and inference scripts and skills. [Start Building With GitHub](https://github.com/nvidia/Cosmos) ### Agent Skills for Synthetic Data Generation Turn coding agents into synthetic data experts for physical AI development. [Try Now](https://brev.nvidia.com/physical-ai) Use Cases ## How Cosmos Accelerates AI Across Industries Use Cosmos WFMs to simulate, reason, and generate data for downstream pipelines in robotics, autonomous vehicles, and industrial vision systems. 1. Robot Learning 2. Autonomous Vehicle Training 3. Video Analytics AI Agents ### Robot Learning Build a robot learning policy that enables embodied agents to operate in real-world environments under both seen and unseen conditions. * Post-train Cosmos 3 on embodiment-specific tasks, environments, camera or sensor layouts, and policies * Run physically accurate closed-loop simulations * Create an end-to-end synthetic data augmentation and evaluation pipeline using agent skills built on Cosmos [See Examples](https://nvidia-cosmos.github.io/cosmos-cookbook/gallery/robotics_inference.html) ### Autonomous Vehicle Training Generate custom, diverse, and high-fidelity sensor data to safely train, test, and validate autonomous vehicles. * Amplify existing data diversity with new weather, lighting, and geolocation data * Post-train to expand into multi-sensor views * Create an end-to-end synthetic data augmentation and evaluation pipeline using agent skills built on Cosmos [See Examples](https://nvidia-cosmos.github.io/cosmos-cookbook/gallery/av_inference.html) ### Video Analytics AI Agents Enhance automation, safety, and operational efficiency across industrial and urban environments. With Cosmos, AI agents can analyze, summarize, and interact with real-time or recorded video streams to: * Deliver real-time contextual alerts * Talk to your videos and extract insights from live camera feeds or large-scale video libraries * Build video analytics AI agents with NVIDIA Metropolis Blueprint for [video search and summarization](https://build.nvidia.com/nvidia/video-search-and-summarization) * Generate synthetic training data to further boost understanding accuracy [Learn More](https://www.nvidia.com/en-us/use-cases/video-analytics-ai-agents.md) Performance ## Runs Best On NVIDIA AI Cosmos 3 is optimized for the best performance on NVIDIA hardware. NVIDIA RTX PRO™ 6000 Blackwell Series Servers accelerate physical AI development for robots, autonomous vehicles, and AI agents across training, synthetic data generation, simulation, and inference. Unlock peak performance for Cosmos world foundation models on NVIDIA Blackwell GB200 for industrial post-training and inference workloads. [Learn More](https://www.nvidia.com/en-us/data-center/rtx-pro-6000-blackwell-server-edition.md) Ecosystem ## Adopted by Leading Physical AI Innovators Model developers from the robotics, autonomous vehicles, and vision AI industries are using Cosmos to accelerate physical AI development. ## Next Steps ### Join the Cosmos Community Connect with Cosmos experts, engage with fellow developers, provide model feedback, and access continued learning through livestreams and recipes. [Join Now](https://discord.com/invite/nvidiaomniverse) #### Cosmos Cookbook A comprehensive guide for working with the NVIDIA Cosmos ecosystem for real-world, domain-specific applications across robotics, simulation, autonomous systems, and physical scene understanding. [Learn More](https://nvidia-cosmos.github.io/cosmos-cookbook/) #### NVIDIA Cosmos Lab NVIDIA Cosmos Lab pioneers large-scale world foundation models that simulate the physical world — the foundation for breakthrough Physical AI. [Explore Now](https://research.nvidia.com/labs/cosmos-lab/cosmos3/) Resources ## The Latest From Cosmos Developers 1. Latest News 2. Sessions 3. Demos [See All Tech Blogs](https://developer.nvidia.com/blog/tag/cosmos/) [See All Topic News](https://blogs.nvidia.com/blog/tag/cosmos/) [See All Press Releases](https://nvidianews.nvidia.com/news/?q=cosmos) [See All](https://www.nvidia.com/en-us/on-demand/playlist/playList-d71fe16e-07d7-4c9c-a34e-e03e3c9a3b4d/) [See All](https://www.youtube.com/playlist?list=PL3jK4xNnlCVc76KbwP4IeLlsFF0F1l7AP) ![YouTube Video](https://img.youtube.com/vi/tGyJO3yR0vU/maxresdefault.jpg) Consent for Optional Cookies (googleCookiePolicyLink)YouTube sets performance, advertising, and other optional cookies(/googleCookiePolicyLink) when you watch embedded videos. To watch this video, you need to turn on optional cookies for the site. By clicking “Accept and Play Video,” you will automatically turn on advertising and other optional cookies for the site and accept our (nvidiaTermsOfServiceLink)Terms of Service(/nvidiaTermsOfServiceLink) (which contains important waivers). Please see our (nvidiaPrivacyPolicyLink)Privacy Policy(/nvidiaPrivacyPolicyLink) and (nvidiaCookiePolicyLink)Cookie Policy(/nvidiaCookiePolicyLink) for more information. Cancel Accept and Play Video Alternatively, you can (youtubeLink)watch this video on YouTube(/youtubeLink). ### Ensuring Safe Autonomous Driving With NVIDIA Halos ![YouTube Video](https://img.youtube.com/vi/GsB7tGB5g-o/maxresdefault.jpg) Consent for Optional Cookies (googleCookiePolicyLink)YouTube sets performance, advertising, and other optional cookies(/googleCookiePolicyLink) when you watch embedded videos. To watch this video, you need to turn on optional cookies for the site. By clicking “Accept and Play Video,” you will automatically turn on advertising and other optional cookies for the site and accept our (nvidiaTermsOfServiceLink)Terms of Service(/nvidiaTermsOfServiceLink) (which contains important waivers). Please see our (nvidiaPrivacyPolicyLink)Privacy Policy(/nvidiaPrivacyPolicyLink) and (nvidiaCookiePolicyLink)Cookie Policy(/nvidiaCookiePolicyLink) for more information. Cancel Accept and Play Video Alternatively, you can (youtubeLink)watch this video on YouTube(/youtubeLink). ### Scaling AV Data With Omniverse and Cosmos ![YouTube Video](https://img.youtube.com/vi/K_Lwzy72jM8/maxresdefault.jpg) Consent for Optional Cookies (googleCookiePolicyLink)YouTube sets performance, advertising, and other optional cookies(/googleCookiePolicyLink) when you watch embedded videos. To watch this video, you need to turn on optional cookies for the site. By clicking “Accept and Play Video,” you will automatically turn on advertising and other optional cookies for the site and accept our (nvidiaTermsOfServiceLink)Terms of Service(/nvidiaTermsOfServiceLink) (which contains important waivers). Please see our (nvidiaPrivacyPolicyLink)Privacy Policy(/nvidiaPrivacyPolicyLink) and (nvidiaCookiePolicyLink)Cookie Policy(/nvidiaCookiePolicyLink) for more information. Cancel Accept and Play Video Alternatively, you can (youtubeLink)watch this video on YouTube(/youtubeLink). ### How Simulation Enables Safer Autonomous Vehicles | Foretellix ![YouTube Video](https://img.youtube.com/vi/HVp2LZ1UUN4/maxresdefault.jpg) Consent for Optional Cookies (googleCookiePolicyLink)YouTube sets performance, advertising, and other optional cookies(/googleCookiePolicyLink) when you watch embedded videos. To watch this video, you need to turn on optional cookies for the site. By clicking “Accept and Play Video,” you will automatically turn on advertising and other optional cookies for the site and accept our (nvidiaTermsOfServiceLink)Terms of Service(/nvidiaTermsOfServiceLink) (which contains important waivers). Please see our (nvidiaPrivacyPolicyLink)Privacy Policy(/nvidiaPrivacyPolicyLink) and (nvidiaCookiePolicyLink)Cookie Policy(/nvidiaCookiePolicyLink) for more information. Cancel Accept and Play Video Alternatively, you can (youtubeLink)watch this video on YouTube(/youtubeLink). ### Accelerating AV Development With NVIDIA Omniverse and Cosmos ![YouTube Video](https://img.youtube.com/vi/S4tvirlG8sQ/maxresdefault.jpg) Consent for Optional Cookies (googleCookiePolicyLink)YouTube sets performance, advertising, and other optional cookies(/googleCookiePolicyLink) when you watch embedded videos. To watch this video, you need to turn on optional cookies for the site. By clicking “Accept and Play Video,” you will automatically turn on advertising and other optional cookies for the site and accept our (nvidiaTermsOfServiceLink)Terms of Service(/nvidiaTermsOfServiceLink) (which contains important waivers). Please see our (nvidiaPrivacyPolicyLink)Privacy Policy(/nvidiaPrivacyPolicyLink) and (nvidiaCookiePolicyLink)Cookie Policy(/nvidiaCookiePolicyLink) for more information. Cancel Accept and Play Video Alternatively, you can (youtubeLink)watch this video on YouTube(/youtubeLink). ### How Robots Learn to Be Robots: Training, Simulation, and Real World Deployment ![YouTube Video](https://img.youtube.com/vi/8Mwrfvq-GeY/maxresdefault.jpg) Consent for Optional Cookies (googleCookiePolicyLink)YouTube sets performance, advertising, and other optional cookies(/googleCookiePolicyLink) when you watch embedded videos. To watch this video, you need to turn on optional cookies for the site. By clicking “Accept and Play Video,” you will automatically turn on advertising and other optional cookies for the site and accept our (nvidiaTermsOfServiceLink)Terms of Service(/nvidiaTermsOfServiceLink) (which contains important waivers). Please see our (nvidiaPrivacyPolicyLink)Privacy Policy(/nvidiaPrivacyPolicyLink) and (nvidiaCookiePolicyLink)Cookie Policy(/nvidiaCookiePolicyLink) for more information. Cancel Accept and Play Video Alternatively, you can (youtubeLink)watch this video on YouTube(/youtubeLink). ### How Robot Brains Dream and Explore Unseen Worlds ![YouTube Video](https://img.youtube.com/vi/9k3Y6KRGPxo/maxresdefault.jpg) Consent for Optional Cookies (googleCookiePolicyLink)YouTube sets performance, advertising, and other optional cookies(/googleCookiePolicyLink) when you watch embedded videos. To watch this video, you need to turn on optional cookies for the site. By clicking “Accept and Play Video,” you will automatically turn on advertising and other optional cookies for the site and accept our (nvidiaTermsOfServiceLink)Terms of Service(/nvidiaTermsOfServiceLink) (which contains important waivers). Please see our (nvidiaPrivacyPolicyLink)Privacy Policy(/nvidiaPrivacyPolicyLink) and (nvidiaCookiePolicyLink)Cookie Policy(/nvidiaCookiePolicyLink) for more information. Cancel Accept and Play Video Alternatively, you can (youtubeLink)watch this video on YouTube(/youtubeLink). ### Build and Test Smart City AI Agents in Digital Twins ![YouTube Video](https://img.youtube.com/vi/9Uch931cDx8/maxresdefault.jpg) Consent for Optional Cookies (googleCookiePolicyLink)YouTube sets performance, advertising, and other optional cookies(/googleCookiePolicyLink) when you watch embedded videos. To watch this video, you need to turn on optional cookies for the site. By clicking “Accept and Play Video,” you will automatically turn on advertising and other optional cookies for the site and accept our (nvidiaTermsOfServiceLink)Terms of Service(/nvidiaTermsOfServiceLink) (which contains important waivers). Please see our (nvidiaPrivacyPolicyLink)Privacy Policy(/nvidiaPrivacyPolicyLink) and (nvidiaCookiePolicyLink)Cookie Policy(/nvidiaCookiePolicyLink) for more information. Cancel Accept and Play Video Alternatively, you can (youtubeLink)watch this video on YouTube(/youtubeLink). ### NVIDIA Cosmos: A World Foundation Model Platform for Physical AI ![YouTube Video](https://img.youtube.com/vi/ZSxYgW-zHiU/maxresdefault.jpg) Consent for Optional Cookies (googleCookiePolicyLink)YouTube sets performance, advertising, and other optional cookies(/googleCookiePolicyLink) when you watch embedded videos. To watch this video, you need to turn on optional cookies for the site. By clicking “Accept and Play Video,” you will automatically turn on advertising and other optional cookies for the site and accept our (nvidiaTermsOfServiceLink)Terms of Service(/nvidiaTermsOfServiceLink) (which contains important waivers). Please see our (nvidiaPrivacyPolicyLink)Privacy Policy(/nvidiaPrivacyPolicyLink) and (nvidiaCookiePolicyLink)Cookie Policy(/nvidiaCookiePolicyLink) for more information. Cancel Accept and Play Video Alternatively, you can (youtubeLink)watch this video on YouTube(/youtubeLink). ### Generating Synthetic Data for Physical AI With NVIDIA Cosmos ![YouTube Video](https://img.youtube.com/vi/zUX-7XgAJN8/maxresdefault.jpg) Consent for Optional Cookies (googleCookiePolicyLink)YouTube sets performance, advertising, and other optional cookies(/googleCookiePolicyLink) when you watch embedded videos. To watch this video, you need to turn on optional cookies for the site. By clicking “Accept and Play Video,” you will automatically turn on advertising and other optional cookies for the site and accept our (nvidiaTermsOfServiceLink)Terms of Service(/nvidiaTermsOfServiceLink) (which contains important waivers). Please see our (nvidiaPrivacyPolicyLink)Privacy Policy(/nvidiaPrivacyPolicyLink) and (nvidiaCookiePolicyLink)Cookie Policy(/nvidiaCookiePolicyLink) for more information. Cancel Accept and Play Video Alternatively, you can (youtubeLink)watch this video on YouTube(/youtubeLink). ### Autonomous Vehicle Simulation With NVIDIA Omniverse and Cosmos ![YouTube Video](https://img.youtube.com/vi/wjCVFfmsai0/maxresdefault.jpg) Consent for Optional Cookies (googleCookiePolicyLink)YouTube sets performance, advertising, and other optional cookies(/googleCookiePolicyLink) when you watch embedded videos. To watch this video, you need to turn on optional cookies for the site. By clicking “Accept and Play Video,” you will automatically turn on advertising and other optional cookies for the site and accept our (nvidiaTermsOfServiceLink)Terms of Service(/nvidiaTermsOfServiceLink) (which contains important waivers). Please see our (nvidiaPrivacyPolicyLink)Privacy Policy(/nvidiaPrivacyPolicyLink) and (nvidiaCookiePolicyLink)Cookie Policy(/nvidiaCookiePolicyLink) for more information. Cancel Accept and Play Video Alternatively, you can (youtubeLink)watch this video on YouTube(/youtubeLink). ### Using NVIDIA Cosmos World Foundation Models for Physical AI Development Previous Next 1. container 1 2. container 2 3. container 3 4. container 3 5. container 3 6. container 3 7. container 3 8. container 3 9. container 3 10. container 3 11. container 3 ## Frequently Asked Questions ### What’s new in NVIDIA Cosmos architecture? Cosmos 3 is built on Mixture of Transformers architecture. Reasoning and generator modules use different transformers for efficient generation and performance. The model therefore reasons first and then generates, resulting in leading physics accuracy across capabilities. Learn more about the architecture [here](https://www.nvidia.com/en-us/glossary/mixture-of-transformers.md). ### What is the licensing model for Cosmos world foundation models? Cosmos WFMs are available under the OpenMDW1.1 license from Linux Foundation. ### How do I post-train Cosmos models for my downstream applications? Cosmos 3 is openly available with post-training scripts on GitHub for each modality and module. In addition, NVIDIA TAO 7 provides a suite of agent skills and tools for fine-tuning vision AI models, including Cosmos 3, with coding agents and natural language prompts. ### Can I build a world model from scratch using tools from the Cosmos platform and my custom or in-house foundation model? Yes, you can leverage Cosmos to build from scratch with your preferred foundation model or model architecture. You can start by using [Cosmos Curator](https://github.com/nvidia-cosmos/cosmos-curate) for video data preprocessing. Then compress and decode your data with [Cosmos tokenizer](https://developer.nvidia.com/blog/state-of-the-art-multimodal-generative-ai-model-development-with-nvidia-nemo/). Once you have processed the data, you can train or fine-tune your model. Using [NVIDIA NIM™ microservices](https://build.nvidia.com/search?q=%22cosmos%22+-nemotron), you can easily integrate your physical AI models into your applications across cloud, data centers, and workstations. You can also use [NVIDIA DGX Cloud](https://www.nvidia.com/en-us/data-center/dgx-cloud/?ncid=pa-srch-goog-542457-DGX-Brand-prsp&_bt=719595959246&_bk=nvidia%20dgx%20cloud&_bm=b&_bn=g&_bg=169380731029&gad_source=1&gclid=EAIaIQobChMI5ofZgLyvigMVlNXCBB0Vsgu8EAAYASAAEgKPbvD_BwE) to train AI models and deploy them at scale anywhere. ### What’s the difference between Cosmos 3 and previous Cosmos models? Cosmos 3 is an omni-model it can generate across text, image, video, sound and action. While Cosmos 2.5 and Cosmos 2 kept perception and generation as separate models and modalities were limited to text, image and video. ### What is the difference between Cosmos and Omniverse? Omniverse creates realistic 3D simulations of real-world tasks by using different generative APIs, SDKs, and NVIDIA RTX rendering technology. Developers can input Omniverse simulations as instructional videos into Cosmos Transfer models to generate controllable, photorealistic synthetic data. Together, Omniverse provides the simulation environment before and after training, while Cosmos provides the foundation models for generating video data and training physical AI models. Learn more about [NVIDIA Omniverse](https://www.nvidia.com/en-us/omniverse.md). NVIDIA [United States](https://www.nvidia.com/object/country-selector.html) * [Privacy Policy](https://www.nvidia.com/en-us/about-nvidia/privacy-policy.md) * [Your Privacy Choices](https://www.nvidia.com/en-us/about-nvidia/privacy-center.md) * [Terms of Service](https://www.nvidia.com/en-us/about-nvidia/terms-of-service.md) * [Accessibility](https://www.nvidia.com/en-us/about-nvidia/accessibility.md) * [Corporate Policies](https://www.nvidia.com/en-us/about-nvidia/company-policies.md) * [Product Security](https://www.nvidia.com/en-us/product-security.md) * [Contact](https://www.nvidia.com/en-us/contact.md) Copyright © 2026 NVIDIA Corporation