On January 27, 2026, Lingbot Technology, the embodied intelligence firm under Ant Group, made an announcement: it was open-sourcing its high-precision spatial perception model, LingBot-Depth. This model is built upon the chip-level raw data sourced from Orbbec's Gemini 330 series binocular 3D cameras. It significantly bolsters environmental depth perception as well as three-dimensional spatial comprehension abilities. By doing so, it delivers more precise and dependable 3D visual assistance to intelligent terminals like robots and autonomous vehicles. In rigorous and authoritative evaluations, this model has demonstrated a remarkable reduction of over 70% in relative error. Additionally, the RMSE error in sparse depth completion tasks has seen a decrease of roughly 47%.
