XPeng and universities release GroundingPI, an open AI model that helps robots see

XPeng and researchers at Peking University, the University of Hong Kong and other universities have released GroundingPI, XPeng said on 4 November. It is an AI model that finds objects in a picture and marks exactly where they are. Robots need that to pick things up. In the team's own tests, its average score across 34 public sets of vision tasks was 73.68, against 71.54 for GPT-6 Astra. Its paper, code and model are public. XPeng says it is research for now, not a plan to put it in its cars or robots.