Lingchu Unveils Embodied Large Model Deployment

创业邦 · wechat · 2026-07-16

This article focuses on Lingchu's demonstrations and partnerships at WAIC, highlighting its embodied large model capabilities rather than just mechanical movements. ### Demonstrations - The booth featured **5 Demos + 1 interactive experience**, covering scenarios like phone box assembly, World Model data conversion, and optical modules - Use cases included insertion detection, vacuum packaging, and automated boxing, emphasizing precision manipulation, long-horizon tasks, and industrial coordination ### Industrial Collaboration - Lingchu partnered with **Changfei Group** to deploy multi-robot collaboration powered by the **Psi-R2 embodied large model** for optical module inspection and packaging - The system achieves sub-millimeter positioning, yield sorting, vacuum sealing, and packing workflows ### Data and Model Strategy - The company invests heavily in human data collection, believing that "model capability determines the upper limit" - They have accumulated **100,000 hours of human data** and **open-sourced a 1,000-hour dataset** - The core logic is driving embodied AI iteration through a "data-model-scenario" closed loop The article concludes that the embodied AI race is about software and iterative capabilities, not just hardware.

Original post →

More from Embodied

Embodied channel →