World Labs has introduced Marble, a commercial world model designed to give artificial intelligence (AI) the ability to perceive, predict and interact with the physical world. The launch signals a shift in the AI landscape as companies move beyond language and image models toward systems that can generate and reason over 3D environments.
Marble is the first product released by World Labs started by Fei Fei Li, which has concentrated its research on spatial intelligence and internal world modeling. The company said the goal is to push AI forward from reading and writing to understanding motion, geometry and cause and effect.
A New Architecture for the Physical World
World Labs describes Marble as a multimodal system that can generate 3D scenes from text prompts, images, videos or spatial sketches, according to its product announcement. The model can extrapolate a single image into a navigable 3D environment and includes an interface called Chisel that lets users draw a rough layout and refine it with natural language.
Marble represents World Labs’ transition from research to commercialization. It is available through freemium and paid tiers that support exports in Gaussian splats, traditional meshes and video files, which allows integration with creative pipelines, simulation tools and real-time rendering engines.
World Labs positions spatial intelligence as the next major layer of AI development. PYMNTS reported that experts view spatial understanding as essential for AI systems that must operate in real-world environments. Most current models work with flat, static data, which limits their usefulness in tasks that require depth, motion or physical reasoning. Marble is designed to bridge that gap by modeling how objects occupy space and interact over time.
Global Firms Are Paying Attention
World Labs enters a relatively emerging field that has expanded quickly over the past year. The South China Morning Post reported that Tencent is expanding its world model efforts, investing in large-scale training runs designed to simulate physical environments, support robotics development and improve generative 3D workflows. The report noted that Tencent views spatial intelligence as a strategic priority and has reorganized parts of its AI research teams to accelerate progress in 3D simulation and multimodal spatial learning.
Tencent’s push reflects a broader trend among major Asian and U.S. companies that are rushing to build systems capable of generating consistent environments, predicting motion and training agentic models with spatial context. These efforts are tied to growth in digital twins, autonomous robotics and immersive applications that depend on accurate physical simulation.
TechCrunch noted that early adopters of Marble include creative studios, simulation developers and teams in gaming and VFX seeking faster ways to build environments. The ability to convert text prompts into structured 3D spaces has drawn interest from organizations that currently rely on manual modeling workflows that are costly and time intensive.
PYMNTS has previously reported that spatial intelligence is also gaining importance in enterprise automation. As companies use AI to support planning, forecasting and real-time decision-making, the ability to understand physical layout and movement improves responsiveness and reliability.
Commercial Potential and Open Questions
Marble’s next test will be its ability to demonstrate value beyond creative production. Expected early interest from entertainment, robotics and simulation sets the stage, but broader enterprise adoption will hinge on accuracy, workflow compatibility and compute efficiency. Spatial intelligence could become a core requirement for future AI agents, but many companies will evaluate performance and integration before committing to 3D-first AI systems.
Source: https://www.pymnts.com/