NVIDIA Launches Cosmos World Foundation Model Platform to Accelerate Physical AI Development
NVIDIA has announced Cosmos, a new platform featuring generative world foundation models (WFMs), tokenizers, and video processing tools designed to advance physical AI systems development, particularly for autonomous vehicles and robots. The platform aims to help developers generate synthetic training data and build custom models more efficiently.
Cosmos WFMs will be available under an open model license, with leading companies including Uber, XPENG, and several robotics firms among early adopters. The platform includes an AI-accelerated data processing pipeline that can process 20 million hours of videos in 14 days using the NVIDIA Blackwell platform, and a new tokenizer offering 8x more compression and 12x faster processing than current solutions.
The platform incorporates trustworthy AI principles with built-in guardrails to mitigate harmful content and includes invisible watermarks for AI-generated content. Cosmos WFMs are now available on Hugging Face and NVIDIA NGC catalog, with optimized NVIDIA NIM microservices coming soon.
NVIDIA ha annunciato Cosmos, una nuova piattaforma che presenta modelli di fondazione generativi per mondi (WFM), tokenizer e strumenti di elaborazione video progettati per avanzare nello sviluppo dei sistemi di intelligenza artificiale fisica, in particolare per veicoli autonomi e robot. La piattaforma ha l'obiettivo di aiutare gli sviluppatori a generare dati di addestramento sintetici e a costruire modelli personalizzati in modo pi霉 efficiente.
I WFM di Cosmos saranno disponibili sotto una licenza di modello aperto, con aziende leader come Uber, XPENG e diverse aziende di robotica tra i primi adottanti. La piattaforma include una pipeline di elaborazione dati accelerata dall'AI che pu貌 elaborare 20 milioni di ore di video in 14 giorni utilizzando la piattaforma NVIDIA Blackwell, e un nuovo tokenizer che offre una compressione 8 volte maggiore e un'elaborazione 12 volte pi霉 veloce rispetto alle soluzioni attuali.
La piattaforma incorpora principi di intelligenza artificiale affidabile con guardrail integrati per mitigare contenuti dannosi e include filigrane invisibili per i contenuti generati dall'AI. I WFM di Cosmos sono ora disponibili su Hugging Face e nel catalogo NVIDIA NGC, con microservizi NVIDIA NIM ottimizzati in arrivo presto.
NVIDIA ha anunciado Cosmos, una nueva plataforma que cuenta con modelos de fundaci贸n generativos para mundos (WFM), tokenizadores y herramientas de procesamiento de video dise帽adas para avanzar en el desarrollo de sistemas de inteligencia artificial f铆sica, particularmente para veh铆culos aut贸nomos y robots. La plataforma tiene como objetivo ayudar a los desarrolladores a generar datos de entrenamiento sint茅ticos y construir modelos personalizados de manera m谩s eficiente.
Los WFM de Cosmos estar谩n disponibles bajo una licencia de modelo abierto, con empresas l铆deres como Uber, XPENG y varias firmas de rob贸tica entre los primeros adoptantes. La plataforma incluye un pipeline de procesamiento de datos acelerado por AI que puede procesar 20 millones de horas de videos en 14 d铆as utilizando la plataforma NVIDIA Blackwell, y un nuevo tokenizador que ofrece una compresi贸n 8 veces mayor y un procesamiento 12 veces m谩s r谩pido que las soluciones actuales.
La plataforma incorpora principios de AI confiable con salvaguardias integradas para mitigar contenido da帽ino e incluye marcas de agua invisibles para el contenido generado por AI. Los WFM de Cosmos ya est谩n disponibles en Hugging Face y en el cat谩logo de NVIDIA NGC, con microservicios NVIDIA NIM optimizados que llegar谩n pronto.
NVIDIA电� Cosmos毳� 氚滍憸頄堨姷雼堧嫟. 鞚� 靸堧鞖� 頂岆灚韽检潃 靸濎劚順� 靹戈硠 旮办磮 氇嵏(WFM), 韱犿伂雮橃澊鞝� 氚� 牍勲敂鞓� 觳橂Μ 霃勱惮毳� 韸轨鞙茧 頃橂┌, 氍茧Μ鞝� AI 鞁滌姢韰� 臧滊皽, 韸鬼瀳 鞛愳湪欤柬枆彀檧 搿滊磭鞚� 氚滌爠鞁滍偆旮� 鞙勴暣 靹り硠霅橃棃鞀惦媹雼�. 鞚� 頂岆灚韽检潃 臧滊皽鞛愱皜 頃╈劚 頉堧牗 雿办澊韯半ゼ 靸濎劚頃橁碃 毵烄钉順� 氇嵏鞚� 氤措嫟 須湪鞝侅溂搿� 甑稌頃樀鐢 雿� 霃勳泙鞚� 欤嫉鐢 瓴冹潉 氇╉憸搿� 頃╇媹雼�.
Cosmos WFM鞚� 鞓ろ攬 氇嵏 霛检澊靹检姢 頃橃棎 鞝滉车霅橂┌, Uber, XPENG 氚� 鞐煬 搿滊磭 須岇偓鞕� 臧欖澊 欤检殧 旮办梾霌れ澊 齑堦赴 靷毄鞛愱皜 霅╇媹雼�. 鞚� 頂岆灚韽检棎电� NVIDIA Blackwell 頂岆灚韽检潉 靷毄頃橃棳 14鞚� 毵岇棎 2000毵� 鞁滉皠鞚� 牍勲敂鞓るゼ 觳橂Μ頃� 靾� 鞛埖鐢 AI 臧�靻� 雿办澊韯� 觳橂Μ 韺岇澊頂勲澕鞚戈臣 順勳灛 靻旊（靺橂炒雼� 8氚� 雿� 毵庫潃 鞎曥稌 氚� 12氚� 雿� 牍犽ジ 觳橂Μ毳� 鞝滉车頃樀鐢 靸堧鞖� 韱犿伂雮橃澊鞝�臧� 韽暔霅橃柎 鞛堨姷雼堧嫟.
鞚� 頂岆灚韽检潃 頃措鞖� 旖橅厫旄犽ゼ 鞕勴檾頃橁赴 鞙勴暅 雮挫灔霅� 臧�霌滊爤鞚检潉 臧栰稑 鞁犽頃� 靾� 鞛埖鐢 AI 鞗愳箼鞚� 韱淀暕頃橁碃, AI搿� 靸濎劚霅� 旖橅厫旄犾棎 雽�頃� 氤挫澊歆� 鞎姷鐢 鞗岉劙毵堩伂毳� 韽暔頃╇媹雼�. Cosmos WFM鞚� 歆�旮� Hugging Face鞕� NVIDIA NGC 旃错儓搿滉犯鞐愳劀 靷毄頃� 靾� 鞛堨溂氅�, 斓滌爜頇旊悳 NVIDIA NIM 毵堨澊韥靹滊箘鞀り皜 瓿� 於滌嫓霅� 鞓堨爼鞛呺媹雼�.
NVIDIA a annonc茅 Cosmos, une nouvelle plateforme comprenant des mod猫les de fondation g茅n茅ratifs pour les mondes (WFM), des tokenizers et des outils de traitement vid茅o con莽us pour faire progresser le d茅veloppement des syst猫mes d'IA physique, en particulier pour les v茅hicules autonomes et les robots. La plateforme vise 脿 aider les d茅veloppeurs 脿 g茅n茅rer des donn茅es d'entra卯nement synth茅tiques et 脿 construire des mod猫les personnalis茅s plus efficacement.
Les WFM de Cosmos seront disponibles sous une licence de mod猫le ouvert, avec des entreprises leaders telles qu'Uber, XPENG et plusieurs entreprises de robotique parmi les premiers adoptants. La plateforme inclut un pipeline de traitement de donn茅es acc茅l茅r茅 par l'IA qui peut traiter 20 millions d'heures de vid茅os en 14 jours en utilisant la plateforme NVIDIA Blackwell, et un nouveau tokenizer offrant une compression 8 fois sup茅rieure et un traitement 12 fois plus rapide que les solutions actuelles.
La plateforme incorpore des principes d'IA fiable avec des garde-fous int茅gr茅s pour att茅nuer les contenus nuisibles et inclut des filigranes invisibles pour les contenus g茅n茅r茅s par l'IA. Les WFM de Cosmos sont d茅sormais disponibles sur Hugging Face et dans le catalogue NVIDIA NGC, avec des microservices NVIDIA NIM optimis茅s qui arriveront bient么t.
NVIDIA hat Cosmos angek眉ndigt, eine neue Plattform, die generative Weltgrundmodelle (WFM), Tokenizer und Videoverarbeitungswerkzeuge umfasst, die f眉r die Entwicklung physischer KI-Systeme, insbesondere f眉r autonome Fahrzeuge und Roboter, entwickelt wurden. Die Plattform zielt darauf ab, Entwicklern zu helfen, synthetische Trainingsdaten zu generieren und ma脽geschneiderte Modelle effizienter zu erstellen.
Die Cosmos WFM werden unter einer offenen Modelllizenz verf眉gbar sein, mit f眉hrenden Unternehmen wie Uber, XPENG und mehreren Robotikunternehmen, die zu den ersten Anwendern geh枚ren. Die Plattform enth盲lt eine KI-beschleunigte Datenverarbeitungspipeline, die in 14 Tagen 20 Millionen Stunden Video mit der NVIDIA Blackwell-Plattform verarbeiten kann, sowie einen neuen Tokenizer, der eine 8-fache Kompression und eine 12-fache schnellere Verarbeitung als aktuelle L枚sungen bietet.
Die Plattform integriert vertrauensw眉rdige KI-Prinzipien mit eingebauten Schutzma脽nahmen zur Minderung sch盲dlicher Inhalte und umfasst unsichtbare Wasserzeichen f眉r von KI generierte Inhalte. Die Cosmos WFM sind jetzt auf Hugging Face und im NVIDIA NGC-Katalog verf眉gbar, mit optimierten NVIDIA NIM-Mikrodiensten, die bald verf眉gbar sein werden.
- Launch of open-source platform that could accelerate AI development in robotics and autonomous vehicles
- Early adoption by major companies including Uber, demonstrating market validation
- Significant performance improvements: 8x more compression and 12x faster processing than current tokenizers
- Dramatic reduction in data processing time: 20M hours of video in 14 days vs 3+ years with CPU-only
- None.
Insights
- New State-of-the-Art Models, Video Tokenizers and an Accelerated Data Processing Pipeline, Optimized for NVIDIA Data Center GPUs, Are Purpose-Built for Developing Robots and Autonomous Vehicles
- First Wave of Open Models Available Now to Developer Community
- Global Physical AI Leaders 1X, Agile Robots, Agility, Figure AI, Foretellix, Uber, Waabi and XPENG Among First to Adopt
LAS VEGAS, Jan. 06, 2025 (GLOBE NEWSWIRE) -- 颁贰厂鈥斅�NVIDIA today announced , a platform comprising state-of-the-art generative , advanced tokenizers, guardrails and an accelerated video processing pipeline built to advance the development of systems such as and .
Physical AI models are costly to develop, and require vast amounts of real-world data and testing. Cosmos world foundation models, or WFMs, offer developers an easy way to generate massive amounts of photoreal, physics-based to train and evaluate their existing models. Developers can also build custom models by fine-tuning Cosmos WFMs.
will be available under an open model license to accelerate the work of the robotics and AV community. Developers can preview the first models on the , or download the family of models and fine-tuning framework from the or .
Leading robotics and automotive companies, including 1X, Agile Robots, Agility, Figure AI, Foretellix, Fourier, , , , , Skild AI, Virtual Incision, Waabi and XPENG, along with ridesharing giant Uber, are among the first to adopt Cosmos.
鈥淭he ChatGPT moment for robotics is coming. Like large language models, world foundation models are fundamental to advancing robot and AV development, yet not all developers have the expertise and resources to train their own,鈥� said Jensen Huang, founder and CEO of NVIDIA. 鈥淲e created Cosmos to democratize physical AI and put general robotics in reach of every developer.鈥�
Open World Foundation Models to Accelerate the Next Wave of AI
NVIDIA Cosmos鈥� suite of open models means developers can the WFMs with datasets, such as video recordings of AV trips or robots navigating a warehouse, according to the needs of their target application.
Cosmos WFMs are purpose-built for physical AI research and development, and can generate physics-based videos from a combination of inputs, like text, image and video, as well as robot sensor or motion data. The models are built for physically based interactions, object permanence, and high-quality generation of simulated industrial environments 鈥� like warehouses or factories 鈥� and of driving environments, including various road conditions.
In his opening , NVIDIA founder and CEO Jensen Huang showcased ways physical AI developers can use Cosmos models, including for:
- Video search and understanding, enabling developers to easily find specific training scenarios, like snowy road conditions or warehouse congestion, from video data.
- Physics-based photoreal synthetic data generation, using Cosmos models to generate photoreal videos from controlled 3D scenarios developed in the 鈩� platform.
- Physical AI model development and evaluation, whether building a custom model on the foundation models, improving the models using Cosmos for reinforcement learning or testing how they perform given a specific simulated scenario.
- Foresight and 鈥渕ultiverse鈥� simulation, using Cosmos and Omniverse to generate every possible future outcome an AI model could take to help it select the best and most accurate path.
Advanced World Model Development Tools
Building physical AI models requires petabytes of video data and tens of thousands of compute hours to process, curate and label that data. To help save enormous costs in data curation, training and model customization, Cosmos features:
- An NVIDIA AI and CUDA庐-accelerated data processing pipeline, powered by , that enables developers to process, curate and label 20 million hours of videos in 14 days using the NVIDIA Blackwell platform, instead of over three years using a CPU-only pipeline.
- , a state-of-the-art visual tokenizer for converting images and videos into tokens. It delivers 8x more total compression and 12x faster processing than today鈥檚 leading tokenizers.
- The framework for highly efficient model training, customization and optimization.
World鈥檚 Largest Physical AI Industries Adopt Cosmos
Pioneers across the physical AI industry are already adopting Cosmos technologies.
1X, an AI and humanoid robot company, launched the dataset using Cosmos Tokenizer. XPENG will use Cosmos to accelerate the development of its humanoid robot. And Hillbot and Skild AI are using Cosmos to fast-track the development of their general-purpose robots.
鈥淒ata scarcity and variability are key challenges to successful learning in robot environments,鈥� said Pras Velagapudi, chief technology officer at Agility. 鈥淐osmos鈥� text-, image- and video-to-world capabilities allow us to generate and augment photorealistic scenarios for a variety of tasks that we can use to train models without needing as much expensive, real-world data capture.鈥�
Transportation leaders are also using Cosmos to build physical AI for AVs:
- Waabi, a company pioneering generative AI for the physical world starting with autonomous vehicles, is evaluating Cosmos in the context of data curation for AV software development and simulation.
- Wayve, which is developing AI foundation models for autonomous driving, is evaluating Cosmos as a tool to search for edge and corner case driving scenarios used for safety and validation.
- AV toolchain provider Foretellix will use Cosmos, alongside , to evaluate and generate high-fidelity testing scenarios and training data at scale.
- Global ridesharing giant Uber is partnering with NVIDIA to accelerate autonomous mobility. Rich driving datasets from Uber, combined with the features of the Cosmos platform and 鈩�, can help AV partners build stronger AI models even more efficiently.
鈥淕enerative AI will power the future of mobility, requiring both rich data and very powerful compute,鈥� said Dara Khosrowshahi, CEO of Uber. 鈥淏y working with NVIDIA, we are confident that we can help supercharge the timeline for safe and scalable autonomous driving solutions for the industry.鈥�
Developing Open, Safe and Responsible AI
NVIDIA Cosmos in line with NVIDIA鈥檚 principles, which prioritize privacy, safety, security, transparency and reducing unwanted bias.
Trustworthy AI is essential for fostering innovation within the developer community and maintaining user trust. NVIDIA is committed to safe and trustworthy AI, in line with the White House鈥檚 voluntary AI commitments and other global AI safety initiatives.
The open Cosmos platform includes guardrails designed to mitigate harmful text and images, and features a tool to enhance text prompts for accuracy. Videos generated with Cosmos and models on the NVIDIA API catalog include invisible watermarks to identify AI-generated content, helping reduce the chances of misinformation and misattribution.
NVIDIA encourages developers to adopt trustworthy AI practices and further enhance guardrail and watermarking solutions for their applications.
Availability
Cosmos WFMs are under NVIDIA鈥檚 open model license on Hugging Face and the NVIDIA NGC catalog. Cosmos models will soon be available as fully optimized microservices.
Developers can access for accelerated video processing and customize their own world models with . offers a fast and easy way to deploy these models, with enterprise support available through the software platform.
NVIDIA also announced new that developers can use for enterprise AI use cases in healthcare, financial services, manufacturing and more.
About NVIDIA
(NASDAQ: NVDA) is the world leader in accelerated computing.
For further information, contact:
Janette Ciborowski
Corporate Communications
NVIDIA Corporation
+1-734-330-8817
jciborowski@nvidia.com
Certain statements in this press release including, but not limited to, statements as to: the benefits, impact, performance and availability of NVIDIA鈥檚 products, services, and technologies, including NVIDIA Cosmos, NVIDIA API catalog, NVIDIA Omniverse platform, NVIDIA AI, NVIDIA CUDA, NVIDIA NeMo Curator, NVIDIA Blackwell platform, NVIDIA Cosmos Tokenizer, NVIDIA NeMo framework, NVIDIA DGX Cloud, and NVIDIA AI Enterprise software platform; third parties adopting NVIDIA鈥檚 products and technologies, and the benefit and impact thereof; and the ChatGPT moment for robotics coming are forward-looking statements that are subject to risks and uncertainties that could cause results to be materially different than expectations. Important factors that could cause actual results to differ materially include: global economic conditions; our reliance on third parties to manufacture, assemble, package and test our products; the impact of technological development and competition; development of new products and technologies or enhancements to our existing product and technologies; market acceptance of our products or our partners' products; design, manufacturing or software defects; changes in consumer preferences or demands; changes in industry standards and interfaces; unexpected loss of performance of our products or technologies when integrated into systems; as well as other factors detailed from time to time in the most recent reports NVIDIA files with the Securities and Exchange Commission, or SEC, including, but not limited to, its annual report on Form 10-K and quarterly reports on Form 10-Q. Copies of reports filed with the SEC are posted on the company's website and are available from NVIDIA without charge. These forward-looking statements are not guarantees of future performance and speak only as of the date hereof, and, except as required by law, NVIDIA disclaims any obligation to update these forward-looking statements to reflect future events or circumstances.
Many of the products and features described herein remain in various stages and will be offered on a when-and-if-available basis. The statements above are not intended to be, and should not be interpreted as a commitment, promise, or legal obligation, and the development, release, and timing of any features or functionalities described for our products is subject to change and remains at the sole discretion of NVIDIA. NVIDIA will have no liability for failure to deliver or delay in the delivery of any of the products, features or functions set forth herein.
漏 2025 NVIDIA Corporation. All rights reserved. NVIDIA, the NVIDIA logo, CUDA, DGX, NGC, NVIDIA Cosmos, NVIDIA NeMo, and NVIDIA Omniverse are trademarks and/or registered trademarks of NVIDIA Corporation in the U.S. and other countries. Other company and product names may be trademarks of the respective companies with which they are associated. Features, pricing, availability and specifications are subject to change without notice.
A photo accompanying this announcement is available at
FAQ
What is NVIDIA Cosmos and when was it launched?
How much faster is NVIDIA Cosmos' data processing compared to traditional methods?
Which major companies are early adopters of NVIDIA Cosmos?
What are the key technical improvements of NVIDIA Cosmos Tokenizer?
Where can developers access NVIDIA Cosmos WFMs?
