Steadvar — News without the noise

Privacy · Terms · About

© 2026 Steadvar. All rights reserved.

Nvidia Unveils Cosmos 3 Open AI World Model for Robots and Autonomous Systems

TechnologyBusiness4h ago
Share

Similar Articles

Open-Weight AI Models Face Growing Safety Risks as Guardrail Removal Tools Proliferate

TechnologySociety1d ago

AI Compute Costs Now Exceeding Employee Salaries in Some Companies

BusinessTechnology4/26/2026

U.S. Cyber Command to Deploy Top AI Models for Cyber Operations

TechnologyPolitics4/28/2026

Anthropic Secures SpaceX Data Center Capacity to Ease AI Compute Constraints

BusinessTechnology5/7/2026

NASA and IBM's Prithvi AI Model Deployed in Orbit for First Time

ScienceTechnology5/7/2026

Nvidia has released Cosmos 3, an open AI model trained on massive multimodal data to help robots and autonomous vehicles understand and predict physical environments. The model generates action data for machines and can simulate rare or dangerous scenarios. Developers can immediately use two versions of the model, with a local 'edge' version coming soon.

Facts First

  • Nvidia unveiled Cosmos 3, an open AI world model designed for robots, autonomous vehicles, and other physical systems.
  • The model was trained on 20 trillion tokens of multimodal data, including nearly a billion images and 400 million videos.
  • Cosmos 3 generates action data like robot joint angles and trajectories to train machines to navigate and manipulate the world.
  • The company is establishing a coalition with initial partners including Agile Robots, Black Forest Labs, and Runway.
  • Two versions are immediately available: a 'super' model for high physics accuracy and a 'nano' model for fast results.

What Happened

Nvidia unveiled Cosmos 3, an open AI world model designed for robots, autonomous vehicles, and other physical systems to understand and predict real-world environments. The model was trained on 20 trillion tokens of multimodal data, which includes nearly a billion images, 400 million real and synthetic videos, ambient audio, text, and action data from humans and robots. Ming-Yu Liu, VP of Nvidia's Cosmos Lab, stated that Cosmos is meant to model how machines move rather than just how scenes look. Developers can use Cosmos 3 to simulate actions in physical environments and build task-specific models for robots and other machines.

Why this Matters to You

This development could accelerate the deployment of more capable and safer robots and autonomous vehicles. By generating rare or dangerous scenarios—like robot collisions or unusual road events—that are difficult to capture in real life, Cosmos 3 may help train these systems to handle unexpected situations, potentially improving public safety. As an open model, it might lower barriers for developers and startups, leading to a wider variety of robotic applications in industries like manufacturing, logistics, and healthcare.

What's Next

Nvidia is immediately releasing two versions of Cosmos 3: a 'super' model for tasks requiring high physics accuracy and a 'nano' model that generates results in fractions of a second. The company announced that an 'edge' model capable of running locally is coming soon. Nvidia is also establishing a coalition of companies to support the effort, with initial partners including Agile Robots, Black Forest Labs, and Runway. The open nature of the model could encourage broader industry adoption and collaboration.

Perspectives

“
Industry Analysts observe that Nvidia is evolving from a chip manufacturer into a foundational platform for physical AI by integrating models and software.
“
Technical Experts explain that world models are essential for enabling AI agents to perform real-world tasks and for helping physical agents become more generalizable through an understanding of the world.
“
Hardware Developers benefit from open models like Cosmos, which allow for easier customization to specific hardware needs and ensure alignment with industry requirements.