World Labs

World Labs is a spatial intelligence platform that develops frontier models capable of understanding, generating, and interacting with 3D environments. Its primary purpose is to provide machine learning systems with a rich, structured understanding of the physical world, enabling more robust perception, planning, and simulation across real and synthetic spaces. By combining 3D scene understanding with generative capabilities, World Labs helps bridge the gap between raw spatial data and actionable, environment-aware intelligence.

The platform ingests multimodal spatial inputs such as 3D scans, point clouds, depth maps, and video, and converts them into consistent, machine-usable scene representations. Its models can reconstruct detailed 3D scenes, infer object relationships, and generate new environments that adhere to real-world geometry and physics constraints. World Labs also supports interactive querying of scenes—such as identifying objects, measuring distances, or reasoning about visibility and navigation—making it suitable for both analysis and simulation workflows. Its focus on scalable, foundation-style spatial models allows organizations to reuse a single core capability across multiple applications, rather than building task-specific pipelines from scratch.