Meituan upgrades "Star Pupil" large model, AI technology achieves second-level blockade of kitchen risks.

On the afternoon of March 11th, it was announced that Meituan has fully upgraded the "Xingmou" field-specific multimodal large model and integrated service system. It is reported that this system was independently developed by Meituan's core local business-intelligent hardware team and was launched in April 2025, with the first public display at the 6th Shanghai Food Safety Forum in December of the same year. After this upgrade, the Meituan "Xingmou" system will significantly improve the high costs, high latency, and poor privacy protection faced by AI technology in food safety scenarios through the new "end-cloud collaboration" and Agent technology architecture, helping the platform's risk oversight to shift from "retrospective tracking" to "real-time blocking". Currently, this system has been piloted in the platform's food safety monitoring field, using technological innovation to ensure "safety on the tip of the consumer's tongue". Up to now, "Xingmou" has conducted 1.96 billion inspections of Meituan's restaurant kitchen environments, issued over 2.4 million warnings, prompting over 50,000 common problems to be rectified by merchants, providing scalable AI capabilities to ensure food safety. As per the plan, Meituan will continue to accelerate the deployment and coverage of the "Xingmou" service system in core business scenarios such as food and retail in 2026. The development team will focus on the complex long-tail challenges in food safety scenarios, further enhancing the active intelligence and cross-scenario transfer abilities of the models and service system.