World Labs unveils Atlas, an omni world model for spatial intelligence, early access soon

theworldlabs · x · 2026-09-02

World Labs introduces Atlas, its next-generation omni world model for spatial intelligence, with early access opening in the coming weeks.

Architecture

A multimodal autoregressive diffusion transformer pretrained from scratch, natively operating on text, images, video, and 3D. All inputs combine into a shared spatial context; generation stays consistent in 3D with everything seen.
Performance improves with training compute, and the team expects this scaling trend to hold.

Four task families

Camera-controlled generation: images and video from one or more images with pixel-perfect camera control, up to 1 minute at 1440p.
Spatial reconstruction: real-world scenes from 1 to dozens of images, outputting novel-view frames and explicit 3D, beating specialized 3D-reconstruction SOTA.
Space-time simulation: models space and time from input video, enabling dramatic reframing and Real-to-Sim workflows for robotics.
Image generation: text-to-image and 360 panoramas, complex prompts, text rendering, varied styles.

Atlas will power future versions of Marble and other products.

More from Models