{"id":677632,"date":"2026-06-01T10:06:11","date_gmt":"2026-06-01T10:06:11","guid":{"rendered":"https:\/\/www.newsbeep.com\/us\/677632\/"},"modified":"2026-06-01T10:06:11","modified_gmt":"2026-06-01T10:06:11","slug":"nvidia-launches-cosmos-3-the-open-frontier-foundation-model-for-physical-ai","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/us\/677632\/","title":{"rendered":"NVIDIA Launches Cosmos 3, the Open Frontier Foundation Model for Physical AI"},"content":{"rendered":"<p style=\"padding-left:0pt;\">News Summary:<\/p>\n<p>&#13;<br \/>\n\tNVIDIA Cosmos 3 is a new leaderboard-topping open physical AI foundation model, built on a breakthrough mixture-of-transformers architecture for physical AI reasoning, world simulation and action generation.&#13;<br \/>\n\tCosmos 3 is the world\u2019s first fully open omnimodel with native vision reasoning and multimodal generation across text, image, video, ambient sound and action for state-of-the-art synthetic data generation and physical AI policy model development.&#13;<br \/>\n\tNVIDIA launches the NVIDIA Cosmos Coalition with leading AI labs and robotics leaders \u2014 including Agile Robots, Black Forest Labs, Generalist, LTX, Runway and Skild AI \u2014 to advance the next generation of open world models.&#13;<\/p>\n<p style=\"padding-left:0pt;\">NVIDIA GTC Taipei\u2014NVIDIA today launched <a href=\"https:\/\/www.nvidia.com\/en-us\/ai\/cosmos\/\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">NVIDIA Cosmos\u2122 3<\/a>, an open world foundation model for physical AI built on a breakthrough <a href=\"https:\/\/www.nvidia.com\/en-us\/glossary\/mixture-of-transformers\/\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">mixture-of-transformers<\/a> architecture that combines vision reasoning, world generation and action prediction in a single system.<\/p>\n<p style=\"padding-left:0pt;\">Cosmos 3 is the world\u2019s first fully open <a href=\"https:\/\/www.nvidia.com\/en-us\/glossary\/omni-model\/\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">omnimodel<\/a> that can natively understand and generate text, images, video, ambient sound and actions with leading physics accuracy, reducing physical AI training and evaluation cycles from months to days.<\/p>\n<p style=\"padding-left:0pt;\">NVIDIA also launched the NVIDIA Cosmos Coalition, a global collaboration between <a href=\"https:\/\/www.nvidia.com\/en-us\/glossary\/world-models\/\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">world model<\/a> builders and AI developers \u2014 including <a href=\"https:\/\/www.agile-robots.com\/en\/news\/detail\/?tx_news_pi1%5Baction%5D=detail&amp;tx_news_pi1%5Bcontroller%5D=News&amp;tx_news_pi1%5Bnews_preview%5D=158&amp;cHash=d40b462afdc783bd58a86c12e0bbb770\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">Agile Robots<\/a>, Black Forest Labs, Generalist, LTX, Runway and Skild AI \u2014 working together to advance next-generation world models.<\/p>\n<p style=\"padding-left:0pt;\">\u201cThe big bang of physical AI is just around the corner thanks to breakthroughs in multimodal reasoning language, vision and world models,\u201d said Jensen Huang, founder and CEO of NVIDIA. \u201cThe Cosmos 3 family of open, frontier omnimodels gives developers a generational leap in ability to build robots, autonomous vehicles and vision AI that perceive, reason, plan and act in the physical world.\u201d<\/p>\n<p style=\"padding-left:0pt;\">A New Architecture for Physical AI<br \/>&#13;<br \/>\nCosmos 3 tackles a fundamental challenge in physical AI: enabling robots, autonomous vehicles (AVs) or vision agents to generalize in the real world with limited training data and fragmented simulation stacks.<\/p>\n<p style=\"padding-left:0pt;\">The model\u2019s mixture-of-transformers architecture pairs a reasoning transformer with an expert generation transformer, enabling Cosmos 3 to understand object interactions, motion and spatial-temporal relationships before generating video and action trajectories.<\/p>\n<p style=\"padding-left:0pt;\">Trained on one of the largest multimodal physical AI datasets \u2014 including billions of samples across text, image, video, sound and action trajectories \u2014 the model gives developers a powerful pretrained foundation for building physical AI systems with less data and lower training costs.<\/p>\n<p style=\"padding-left:0pt;\">Developers can use Cosmos 3 as:<\/p>\n<p>&#13;<br \/>\n\tA <a href=\"https:\/\/www.nvidia.com\/en-us\/glossary\/vision-language-models\/\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">vision language model<\/a> that understands and reasons across modalities.&#13;<br \/>\n\tA world model or video foundation model that simulates physical environments and predicts future world states for training and evaluation.&#13;<br \/>\n\tThe backbone for <a href=\"https:\/\/www.nvidia.com\/en-us\/glossary\/world-action-model\/\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">world action models<\/a> that help train robots to perform specific tasks.&#13;<\/p>\n<p style=\"padding-left:0pt;\">Cosmos 3 models deliver leading results on physical AI benchmarks. Among open models, it ranks first across <a href=\"https:\/\/artificialanalysis.ai\/leaderboards\/models?weights=open&amp;reasoning=reasoning&amp;size=small%2Cmedium\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">Artificial Analysis<\/a>, <a href=\"https:\/\/physics-iq.github.io\/\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">Physics-IQ<\/a>, <a href=\"https:\/\/huggingface.co\/spaces\/shi-labs\/physical-ai-bench-leaderboard\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">PAI-Bench<\/a> and <a href=\"https:\/\/huggingface.co\/spaces\/DAGroup-PKU\/RBench-Leaderboard\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">R-Bench<\/a> for world generation accuracy, <a href=\"https:\/\/research.nvidia.com\/labs\/srl\/projects\/robolab\/\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">RoboLab<\/a> and <a href=\"https:\/\/robo-arena.github.io\/leaderboard\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">RoboArena<\/a> for action policy, and the <a href=\"https:\/\/huggingface.co\/spaces\/clemson-computing\/VANTAGE-Bench-Leaderboard\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">VANTAGE-Bench<\/a> and <a href=\"https:\/\/eval.aicitychallenge.org\/aicity2026\/submission\/leaderboard?trackId=3&amp;type=general\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">TAR<\/a> leaderboards for vision understanding.<\/p>\n<p align=\"left\" style=\"padding-left:0pt;\">The Cosmos 3 lineup gives developers options for different stages of physical AI development:<\/p>\n<p>&#13;<br \/>\n\tCosmos 3 Super for post-training robotics and AV models that need the highest physics accuracy and generation quality.&#13;<br \/>\n\tCosmos 3 Nano for high-quality video and action reasoning in fractions of a second.&#13;<br \/>\n\tCosmos 3 Edge, coming soon, for real-time inference at the edge.&#13;<\/p>\n<p style=\"padding-left:0pt;\">Cosmos Coalition Accelerates Open World Model Development<br \/>&#13;<br \/>\nThe Cosmos Coalition is a global collaboration between world model builders, AI developers and physical AI leaders to advance open world models across industries, enabling members to contribute models, research and evaluation techniques while using Cosmos 3 technologies, training tools and <a href=\"https:\/\/www.nvidia.com\/en-us\/data-center\/dgx-cloud\/\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">NVIDIA DGX\u2122 Cloud<\/a> infrastructure for large-scale training.<\/p>\n<p style=\"padding-left:0pt;\">Founding coalition members include Agile Robots, Black Forest Labs, Generalist, LTX, Runway and Skild AI. By building in the open and contributing across a shared ecosystem, the coalition aims to enable faster innovation, broader interoperability and more rapid advances in physical AI.<\/p>\n<p style=\"padding-left:0pt;\">Developers Build on Cosmos<br \/>&#13;<br \/>\nThe Cosmos platform powers NVIDIA\u2019s physical AI stack to accelerate training and evaluation workflows across industries. The platform now includes new datasets for robotics, physics, human motion, autonomous driving, warehouse safety and spatial reasoning, as well as new <a href=\"https:\/\/nvidianews.nvidia.com\/news\/nvidia-releases-major-collection-of-open-source-agent-tools-and-skills-for-physical-ai\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">physical AI agent skills<\/a> for neural scene reconstruction, defect-image generation and video augmentation.<\/p>\n<p style=\"padding-left:0pt;\">Physical AI developers are building on the Cosmos platform across industries \u2014\u00a0Agile Robots, Doosan Robotics, LG Electronics, Samsung and Skild AI for robotics, Li Auto for AVs, and Centific, <a href=\"https:\/\/fogsphere.com\/fogsphere-announces-cosmos-3-support\/\" rel=\"nofollow noopener\" target=\"_blank\" title=\"Fogsphere\">Fogsphere<\/a>, <a href=\"https:\/\/www.linkervision.com\/post\/linker-vision-unveils-application-driven-ai-grid-for-agentic-video-reasoning-at-scale\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">Linker Vision<\/a>, <a href=\"https:\/\/www.milestonesys.com\/resources\/content\/articles\/milestone-hafnia-nvidia-cosmos-3\/\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">Milestone Systems<\/a> and Yuan for <a href=\"https:\/\/www.nvidia.com\/en-us\/use-cases\/video-analytics-ai-agents\/\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">vision AI agents<\/a> to power industrial AI and smart spaces applications.<\/p>\n<p style=\"padding-left:0pt;\">Availability<br \/>&#13;<br \/>\nCosmos 3 Super and Cosmos 3 Nano are available now, with Cosmos 3 Edge coming soon for real-time inference. Developers can try Cosmos 3 on <a href=\"https:\/\/build.nvidia.com\/models?q=cosmos\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">build.nvidia.com<\/a>, download open models from <a href=\"https:\/\/huggingface.co\/collections\/nvidia\/cosmos3\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">Hugging Face<\/a>, customize models and generate synthetic data with Hugging Face Diffusers and resources on <a href=\"https:\/\/github.com\/nvidia\/Cosmos\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">GitHub<\/a>, and deploy the models as <a href=\"https:\/\/www.nvidia.com\/en-us\/ai-data-science\/products\/nim-microservices\/\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">NVIDIA NIM<\/a>\u2122 microservices.<\/p>\n<p style=\"padding-left:0pt;\">Model builders and software providers can accelerate access, customization and deployment of Cosmos for key reasoning and synthetic data generation workloads using physical AI agent skills on GitHub through inference services and cloud infrastructure partners including <a href=\"https:\/\/www.baseten.co\/blog\/nvidia-cosmos-3-robots-finally-take-over\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">Baseten<\/a>, <a href=\"https:\/\/wandb.ai\/wandb_fc\/nvidia-cosmos\/reports\/Build-a-Robotics-Data-Flywheel-on-CoreWeave-using-NVIDIA-Cosmos-3--VmlldzoxNzA3MzA2Ng\" rel=\"nofollow noopener\" target=\"_blank\" title=\"CoreWeave\">CoreWeave<\/a>, Microsoft Azure, <a href=\"https:\/\/nebius.com\/blog\/posts\/run-physical-ai-workflows-not-glue-code%20Classmethod:%20https:\/\/dev.classmethod.jp\/articles\/dgx-spark-cosmos3-omni-world-model-policy\/\" rel=\"nofollow noopener\" target=\"_blank\" title=\"Nebius\">Nebius<\/a>, Deep Infra and Classmethod.<\/p>\n<p style=\"padding-left:0pt;\">Watch the <a href=\"https:\/\/www.nvidia.com\/en-tw\/gtc\/taipei\/keynote\/\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">keynote<\/a> from Huang, learn more at <a href=\"https:\/\/www.nvidia.com\/en-tw\/gtc\/taipei\/\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">NVIDIA GTC Taipei<\/a> and explore these <a href=\"https:\/\/www.nvidia.com\/en-tw\/gtc\/taipei\/session-catalog\/?tab.catalogallsessionstab=16566177511100015Kus&amp;search=STW61026%2C%20STW61028%2C%20STW61011%2C%20STW61066%2C%20STW61024%2C%20STW61062%2C%20STW61036#\/\" rel=\"nofollow noopener\" target=\"_blank\" title=\"\">physical AI sessions<\/a>.<\/p>\n","protected":false},"excerpt":{"rendered":"News Summary: &#13; NVIDIA Cosmos 3 is a new leaderboard-topping open physical AI foundation model, built on a&hellip;\n","protected":false},"author":2,"featured_media":677633,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[27],"tags":[28],"class_list":["post-677632","post","type-post","status-publish","format-standard","has-post-thumbnail","category-business","tag-business"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/posts\/677632","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/comments?post=677632"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/posts\/677632\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/media\/677633"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/media?parent=677632"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/categories?post=677632"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/tags?post=677632"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}