World Labs Unveils Atlas World Model

World Labs posted about Atlas, an autoregressive multimodal model for spatial intelligence. Company statements describe it as generating image and video frames with pixel-perfect camera control while outputting explicit 3D reconstructions from one or many input images. Researcher Gowthami Somepalli and others at the firm noted its next-frame prediction design, which supports camera-directed video, novel view synthesis, and sparse 3D reconstruction. Posts state the model was pretrained from scratch on multimodal inputs and improves with additional images. A linked blog on worldlabs.ai provides further details. Replies from external researchers and investors acknowledged the release.