Ant Group Unveils Lingbo World Model for Embodied AI
Ant Group Launches Lingbo World Model
Ant Group has unveiled Lingbo World Model, a breakthrough [embodied AI](/glossary/embodied-ai) platform that creates physics-accurate digital twins of real-world environments by combining sensor data with massive internet knowledge.
Dual Data Approach
The Lingbo system employs a unique dual-data strategy: real-world physical data collected from sensors across industrial, commercial, and residential scenarios, and internet-scale knowledge data extracted from billions of documents, images, and videos. This hybrid approach enables the model to understand both the physical laws governing the world and the semantic context of objects and interactions.
Physics Engine Integration
At the core of Lingbo lies a proprietary physics simulation engine that accurately models rigid body dynamics, soft tissue deformation, fluid mechanics, and thermal transfer. Unlike traditional game engines optimized for visual realism, Lingbo prioritizes physical accuracy for robot training and simulation-to-real transfer.
Claw Harness Robot Platform
Alongside the world model, Ant Group demonstrated the Claw Harness [dexterous manipulation](/glossary/dexterous-manipulation) platform, featuring a 7-degree-of-freedom robotic arm with 12-axis force-torque sensing. The system has already achieved 99.2% success rate in pick-and-place tasks across 500+ object categories in controlled factory environments.
Tutu Quadruped Guide Robot
The Tutu quadruped robot represents Lingbo's first commercial deployment in public service. Designed for visually impaired navigation assistance, Tutu combines LiDAR [SLAM](/glossary/slam) with visual-language understanding to detect obstacles, recognize crosswalks, and provide audio guidance. Field trials in Hangzhou and Shanghai have shown 94% user satisfaction rates.
Full-Stack [Embodied Intelligence](/glossary/embodied-ai)
Ant Group's announcement positions it among China's leading full-stack [embodied AI](/glossary/embodied-ai) providers, joining companies like Unitree, Agibot, and DeepWay. The company plans to open-source portions of the Lingbo simulation engine through its developer platform, with commercial API access scheduled for Q3 2026.
Industrial Deployment Trajectory
Initial deployments focus on 3C electronics manufacturing and warehouse logistics, where Lingbo's digital twins reduce robot training time by 70% compared to traditional methods. Long-term applications include smart cities, disaster response, and home service robotics.
