As the echoes of Meituan's lower-than-anticipated operating profits in the second quarter of 2025 still reverberate, the company, on November 10th, officially rolled out the public beta of its inaugural AI IDE product, Meituan AI, despite the prevailing profit pressures. This move is perceived as a 'strategic offensive' by the local services behemoth, with its 'Little Cat Paw' finally unsheathing its sharp claws.

The AI Capital Bureau posits that this tool, boasting a 95% internal R&D penetration rate and over 50% AI-generated incremental code, is not a mere isolated technological foray but a pivotal juncture in Meituan's efficiency revolution and growth logic reconfiguration under the weight of performance pressures. Against the backdrop of a deeply entrenched perception of its 'food delivery' business, Meituan has quietly constructed an AI system characterized by 'a robust bottom-layer model foundation, business scenario breakthroughs, and internal-external ecosystem linkage' through substantial R&D investments. The unveiling of CatPaw is gradually unveiling the performance value inherent in this system.

The product evolution trajectory of CatPaw mirrors Meituan's core AI strategy: 'internal value verification preceding external capability output.' Since 2023, the tool has undergone internal testing as an editor plugin, facilitating full-project linkage analysis through Agent collaboration and alleviating repetitive labor intensity for R&D personnel with features such as real-time code completion and batch modification prediction.

This 'internal-first, external-later' approach resonates with Meituan's rhythm in deploying multiple technologies. According to Meituan, CatPaw has significantly curtailed repetitive coding time for R&D teams during its internal application phase. Its self-developed LongCat programming large model further elucidates Meituan's 'scenario-specific training' technical route—diverging from some companies' pursuit of parameter races in general-purpose large models, Meituan zeroes in on cultivating targeted AI capabilities within vertical domains.

If CatPaw is viewed as the 'front-end application carrier' of Meituan's AI technology, the LongCat series of large models serves as its underlying technical bedrock. From September to November 2025, Meituan unveiled four LongCat models within a span of three months, encompassing language understanding, video generation, reasoning computation, and full-modal interaction, thereby forming a technology matrix adaptable to diverse scenarios. From a technical architecture standpoint, the LongCat series adopts a MoE (Mixture of Experts) architecture, achieving efficient operation with 27 billion activated parameters out of a total of 560 billion, matching the immediacy demands of local services with millisecond-level response speeds.

Contrary to the general-purpose large model routes pursued by companies like Alibaba and Baidu, Meituan trains its models using vertical domain data amassed over a decade, including dining transactions, user behavior, and merchant operations, enabling differentiated precision in specific scenarios such as 'site selection and pricing' and 'order scheduling.' This 'data-model-business' closed loop constitutes a distinctive hallmark of Meituan's AI but also confronts challenges in terms of general-purpose capabilities.

The breadth of technology deployment serves as a crucial metric for gauging the depth of AI strategic layout. Currently, Meituan has forged an AI product matrix encompassing B-end merchants, C-end users, and internal operations: B-end products like 'Kangaroo Consultant' offer site selection decision references for merchants, 'Kangaroo Housekeeper' handles fundamental operations for food delivery stores, and 'Smart Shopkeeper' optimizes in-store dining management processes; C-end smart assistant 'Xiaomei' supports automated services such as restaurant reservations and order inquiries; in core delivery scenarios, the intelligent decision-making platform facilitates daily scheduling of 60 million orders, while drones and autonomous delivery vehicles extend AI from the 'decision layer' to the 'fulfillment layer.'

From a business synergy perspective, these products have forged initial linkages. For instance, after a C-end user completes a restaurant reservation via 'Xiaomei,' the B-end 'Smart Shopkeeper' can initiate an AI-powered confirmation call process, constructing a 'demand-response-execution' chain. Nevertheless, the penetration rate and actual effectiveness of these AI tools still necessitate longer-term market validation, and the service threshold for small and medium-sized merchants requires further lowering.

Against the backdrop of overall profit pressures in the internet industry, Meituan's sustained AI investment stands as an industry exemplar.

The second-quarter 2025 financial report reveals that Meituan achieved RMB 91.84 billion in revenue in the second quarter of 2025, marking a year-on-year increase of 11.7%, with an adjusted net profit of RMB 1.49 billion, down 89% year-on-year and significantly below market expectations. However, Meituan's R&D expenditure surged by 17.2% year-on-year, with a notable uptick in the proportion of investment in AI infrastructure and model R&D.

Behind this 'counter-cyclical investment' lies Meituan's astute judgment on the shifting competitive logic in the local services industry—from price competition in incremental markets to efficiency competition in stock markets, with AI perceived as the core driver for efficiency enhancement.

Judging from the current progress, CatPaw's high internal penetration, the GMV boost for certain merchants from AI tools at the merchant end, and the open-source attempts of the LongCat model indicate that Meituan is exploring an AI development path of 'endogenous growth-external empowerment-ecosystem co-construction,' which shares certain parallels with the AI strategies of companies like Tencent and ByteDance.

The public beta of CatPaw is not merely a public display of Meituan's AI technology but also a phased manifestation of its 'retail + technology' strategic transformation. Compared to the general-purpose AI strategic layout of internet giants, Meituan's AI is characterized by deep integration with local services scenarios, defining technological directions based on specific business needs and validating technological value through tangible business outcomes.

It can be argued that when the computational power support of the LongCat model converges with the operational data of millions of merchants, and when the R&D efficiency improvements of CatPaw align with the intelligent scheduling of food delivery, Meituan is endeavoring to construct an AI-driven local services ecosystem, which may underpin a rebound in its performance.

CITIC Construction Investment Securities research posits that Meituan's food delivery business is projected to incur losses exceeding RMB 17 billion in the third quarter, with losses peaking in this quarter. Since the fourth quarter, competition intensity has abated, subsidies have gradually returned to rationality, and Meituan's food delivery order market share has rebounded, with food delivery UE losses expected to halve.

For the industrial application of AI, Meituan's practice also offers a 'scenario-first' reference sample, whose value lies in demonstrating that the commercial potential of AI technology is not solely reflected in technological parameter breakthroughs but also in its adaptability to industrial needs, providing a feasible path for the industrial application of AI technology.

AI Capital Bureau is a professional observation and analysis platform dedicated to capital market dynamics in the artificial intelligence realm. We meticulously track capital operations such as financing, listings, and mergers and acquisitions of AI companies, conduct in-depth analysis of industry trends and investment opportunities, and furnish valuable insights for industry participants. In an era marked by rapid AI technological development and profound capital integration, we are committed to serving as a bridge connecting AI innovation and the capital market, aiding Chinese AI companies in realizing value discovery and growth.

Risk Warning and Disclaimer: The market is fraught with risks, and investment should be approached with caution. This article does not constitute investment advice and should not be utilized as practical operational guidance. Trading risks are borne solely by the investor.