World Labs, founded by researcher Fei-Fei Li, launched a multimodal 3D world model named Atlas.

The model generates image and video frames at resolutions up to 1440p.

Atlas supports video durations of up to one minute with precise camera control.

The system processes text, images, video, and 3D data within a unified spatial framework.

This release advances the company’s goal of developing spatial intelligence for physical world interaction.

Atlas succeeds the company’s previous 3D generation model, Marble.

World Labs has raised over $1.2 billion from investors including Autodesk, Nvidia, AMD, and Cisco.