World Labs

American AI company specializing in spatial intelligence and large world models. Founded by Fei-Fei Li in 2024. Headquarters: San Francisco, California.

Events

World Labs Founded by Fei-Fei Li

AI pioneer Fei-Fei Li co-founded World Labs with Justin Johnson, Christoph Lassner, and Ben Mildenhall in San Francisco. The company aimed to build frontier 'large world models' for 3D spatial intelligence — AI that can perceive, generate, and interact with the physical world.

Marble Beta Preview Released

World Labs opened Marble, its first commercial product, as a limited beta preview. Marble is a generative multimodal world model that creates persistent, navigable 3D worlds from text, images, or video — one of the first commercial world model products.

Marble Generally Available Launch

World Labs made Marble generally available to everyone at marble.worldlabs.ai. Described as a 'first-in-class generative multimodal world model,' Marble lets users create, edit, and share high-fidelity 3D worlds. The launch marked World Labs as one of the first companies to commercialize world model technology.

World Labs Raises $1 Billion

World Labs raised $1 billion in a funding round with $200M from Autodesk, plus participation from AMD, NVIDIA, Fidelity, and others. The round accelerated the company's spatial intelligence efforts and expanded its integration into 3D creative workflows.

World Labs Acquires SceniX

World Labs acquired SceniX, a robotics simulation startup founded by Yunzhu Li and Changxi Zheng. SceniX built simulation platforms to close the real-to-sim gap — the mismatch between virtual training environments and physical hardware. The acquisition bridged World Labs' generative world models with embodied AI and robotics.

World Labs Introduces Atlas, an Omni World Model for Spatial Intelligence

World Labs Introduces Atlas, an Omni World Model for Spatial Intelligence

World Labs introduced Atlas, a next-generation omni world model pretrained from scratch to natively operate on text, images, video, and 3D via a multimodal autoregressive diffusion transformer. Atlas performs camera-controlled generation (up to one minute of 1440p video), spatial reconstruction outperforming specialized 3D models, and space-time simulation enabling real-to-sim workflows for robotics. The launch marked the company's transition from its Marble product to a general-purpose world-model platform aimed at robotics and simulation.