Meituan, a leading lifestyle service platform in China, has unveiled a next-generation artificial intelligence (AI) large model designed to support small business operations. The model aims to enhance the competitiveness of existing lifestyle service platforms by utilizing AI for revenue analysis, inventory management, and marketing strategy development.
On November 2, Meituan announced the launch of its next-generation large model, LongCat-2.5, as reported by the state-run Xinhua News Agency on November 4. The model boasts a total of 1.6 trillion parameters and is a native multimodal model specialized for performing complex tasks over extended periods. Notably, the entire training process for the model was conducted using Chinese computing power.
This model is an upgrade from LongCat-2.0, which was released in June. It enhances multimodal capabilities for understanding visual information and improves software development capabilities, allowing the AI agent to perform complex tasks from start to finish.
Meituan's ultimate goal in developing LongCat is not to sell the AI model itself but to integrate it into its core business operations. By leveraging the vast data accumulated from food delivery, dining, and hotel booking services, Meituan aims to improve the operational efficiency of its franchise partners and strengthen platform competitiveness.
Wang Xing, CEO of Meituan, stated during a conference call on August 28 regarding the second-quarter earnings report, "We will not become a token factory. Our models and AI products will be used to support our core business and enhance the experiences of users and franchise partners, as well as improve the company's operational efficiency."
Meituan has also launched an AI management platform called Miaoshu, based on the LongCat series. Miaoshu assists lifestyle service providers, such as restaurants, in understanding industry-specific management metrics and brand operation rules, enabling them to perform actual business tasks. It can call relevant tools and systems to handle management tasks within the approved scope of franchise partners.
Real-world applications have already been demonstrated. LongCat-2.5 analyzed multi-channel management data for a brand, identifying that while the gross merchandise value (GMV) increased by 23.2%, channel profits decreased by 1.7%. It subsequently provided resource allocation strategies for each channel by comprehensively analyzing inventory costs, customer repurchase rates, and inventory risks. This goes beyond simple data summarization to offer management diagnostics and decision-making support. Currently, Miaoshu is being utilized for management analysis, store management, and customer acquisition tasks by various Chinese franchise brands, including Guming, Xushengji, and Yihetang.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.