Ant Group's Robbyant Open-Sources LingBot-Map for Real-Time 3D Reconstruction
Event summary
- Robbyant, Ant Group's embodied AI unit, open-sourced LingBot-Map, a streaming 3D reconstruction model on April 16, 2026.
- LingBot-Map achieves an Absolute Trajectory Error (ATE) of 6.42 meters on the Oxford Spires dataset, a near 2.8x improvement over prior streaming methods.
- The model operates in real-time at 20 FPS and maintains accuracy over long video sequences exceeding 10,000 frames.
- LingBot-Map is part of Robbyant's growing suite of embodied AI models, including LingBot-Depth, LingBot-VLA, LingBot-World, and LingBot-VA.
The big picture
Robbyant's LingBot-Map represents a significant advancement in real-time spatial understanding, addressing the core challenge of balancing geometric accuracy, temporal consistency, and computational efficiency. This development aligns with the broader industry trend toward embodied AI, where robots and autonomous systems require sophisticated perception capabilities. Ant Group's strategic focus on open-sourcing foundational models positions Robbyant as a key player in the evolving landscape of intelligent devices and services.
What we're watching
- Adoption Pace
- How quickly LingBot-Map will be integrated into robotics, autonomous vehicles, and AR devices.
- Competitive Edge
- Whether Robbyant's open-sourcing strategy will solidify its lead in embodied AI.
- Technical Scalability
- The pace at which LingBot-Map can scale to handle more complex environments and longer sequences.
Related topics