World Labs, founded by AI pioneer Fei-Fei Li, announced the general release of Marble—a world model capable of generating persistent 3D environments from text, image, and video prompts. The platform represents one of the first commercially available “spatial intelligence” models, bridging the gap between LLMs and physical world modeling. Official announcement: World Labs Blog.

• Creative flexibility: Users can generate, edit, and expand 3D worlds via multimodal inputs, combining text, image, and video prompts.
• Export options: Supports export as Gaussian splats, meshes, or videos—integrating with gaming, VFX, and VR pipelines.
• Launch details: After a September preview, Marble is now fully available with freemium and paid plans starting at $20/month.
• Research context: The release aligns with Li’s recent essay on “spatial intelligence,” arguing that world models are the next logical evolution beyond text-based LLMs.

Why it matters: Marble signals the rise of world models—AI systems that understand and generate within 3D space rather than text alone. From robotics and simulation to cinematic design and architecture, this technology could become foundational for how AI perceives, builds, and interacts with the physical world.