Brain
World Labs' first commercial product — a multimodal world model that generates spatially consistent, high-fidelity, persistent 3D worlds from a single image, video, or text prompt. Users can move through, edit, and inhabit the generated worlds. Spatial intelligence foundation for embodied AI, robotics, and simulation.
World model · Maturity: Commercial · Closed
Machine-readable surfaces
- Markdown mirror: /brains/marble.md
- RSS feed: /brains/marble/feed.xml
- JSON-LD: embedded in this page’s head
- REST API: /v1/brains/4bcc0ba0-e1e9-42c9-b6d0-2ca4b256b37b
- Revision history: /brains/marble/history
- Data documentation: /data
- Query this programmatically: Deploy MCP
Architecture
Large World Model (LWM) for spatial intelligence and 3D world generation
Key facts
- General availability (primary World Labs)
- World Labs' Nov 12 2025 post says Marble, its frontier multimodal world model, is generally available after a prior beta preview. Marble can create 3D worlds from text, images, video, or coarse 3D layouts, and supports interactive edit, expand, and combine workflows. Generated worlds can be exported as Gaussian splats, meshes, or videos. Product URL cited is marble.worldlabs.ai (World Labs)
- Chisel and expansion tools (primary World Labs)
- The same post introduces Chisel for sculpting coarse 3D layouts that Marble then styles from a text prompt, plus one-step world expansion and composer mode for combining worlds. Frame capability list as company product description at launch, not a robotics fleet deployment (World Labs)
- Marble developer documentation
- World Labs provides developer documentation for Marble and its world-generation API, extending the product record beyond the launch announcement (World Labs Marble documentation)
- Persistent 3D world generation
- World Labs describes Marble as a multimodal world model that generates spatially consistent, persistent 3D worlds from text, images, video, or coarse 3D layouts (Marble world model).
- World export formats
- World Labs says Marble outputs can be exported as Gaussian splats, meshes, or videos, making the product relevant to simulation and embodied-AI workflows (Marble world model).
- World generation API
- World Labs documents a Marble World API endpoint for generating worlds from text prompts and identifies Marble 1.1 and Marble 1.1 Plus as generation models (Marble API).
Developed by (1)
Common questions
What is Marble?
World Labs' first commercial product — a multimodal world model that generates spatially consistent, high-fidelity, persistent 3D worlds from a single image, video, or text prompt. Users can move through, edit, and inhabit the generated worlds. Spatial intelligence foundation for embodied AI, robotics, and simulation.
Who developed Marble?
Marble is credited to World Labs on the DEPLOY registry. Each developer attribution is verified via primary sources.
Is Marble open source?
No. Marble is recorded as proprietary on the DEPLOY registry. Model weights and source are not publicly available.
What type of AI is Marble?
Marble is a world model, built on a Large World Model (LWM) for spatial intelligence and 3D world generation architecture on the DEPLOY registry.
What is Marble's maturity stage?
Marble is at the commercial stage on the DEPLOY maturity ladder. Commercial stage means production-grade deployments are operating at named customer sites.
Which robots run on Marble?
No robot models on the DEPLOY registry are recorded as running Marble. DEPLOY wires brain-to-model connections only when the wiring is verifiable from primary sources; absence may reflect pre-deployment or unverified manufacturer claims.
Sources (3)
Methodology: Verified · 3 sources (no primary) · last reviewed 2026-10-09
Verification posture
Verified
Low confidence
Review state
Stable
Last reviewed 2026-10-09
Maturity + lifecycle
Maturity stage: commercial
Sources by quality tier
- 3
- unclassified
- Unclassified source
The framework is documented at /methodology. Corrections at /corrections. Reviewer: DEPLOY editorial team.
Methodology surface for Marble.Canonical ID 4bcc0ba0-e1e9-42c9-b6d0-2ca4b256b37b