A multimodal world model in select-partner early access for camera-controlled generation, spatial reconstruction, and simulation from text, images, video, and 3D.
World Labs introduced Atlas in September 2026 as a step toward spatial intelligence. The model was pretrained from scratch to work with text, images, video and 3D information in a shared spatial context. That approach supports several connected tasks: generating camera-controlled views, reconstructing scenes and simulating changes through space and time. Rather than treating each image as an isolated output, Atlas aims to keep new views consistent with the world it has already seen. World Labs described it as a foundation for future versions of Marble and other products.
Be the first Cosign
Add it to your Products I use list. You can also share why you use it.
