World Labs has officially launched Marble, a generative multimodal world model that constructs 3D worlds from text, images, videos, or basic 3D layouts. This release marks the transition from a limited beta, where only selected users could experiment, to public access. The announcement targets creators across industries such as gaming, VFX, design, and robotics, as well as technical professionals and artists eager to experiment with cutting-edge 3D AI modeling.
Introducing Marble by World Labs: a foundation for a spatially intelligent future.
— World Labs (@theworldlabs) November 12, 2025
Create your world at https://t.co/V267VJu1H9 pic.twitter.com/T00mtETmCA
Marble distinguishes itself with its ability to accept a wide range of input types and provide iterative editing, granular world expansion, and composition features. Users can export their creations as Gaussian splats, high-fidelity or collider meshes, and videos, allowing integration with established 3D workflows. The new Chisel tool further addresses advanced use cases, enabling direct sculpting and integration of existing 3D assets. Compared to previous generative 3D solutions, Marble's multimodal, compositional, and export options offer more detailed control and broader compatibility with industry standards.
The company behind Marble, World Labs, has focused on bringing spatial intelligence to artificial intelligence through world models. With this general availability, World Labs is positioning Marble as a platform for both creative professionals and technical innovators to push the boundaries of 3D creation. Early users and industry insiders have noted the platform's flexibility and practical export capabilities, highlighting its potential to impact multiple sectors.
