中国GenAI市场洞察：企业级大模型调用全景研究

Investment Rating - The report does not explicitly state an investment rating for the industry Core Insights - The Chinese enterprise-level GenAI market is experiencing explosive growth, with daily model invocation reaching 101,865 billion tokens in the first half of 2025, a 363% increase from 21,999 billion tokens in the second half of 2024 [8][18][11] - The market is transitioning towards a dual-track development of open-source and closed-source models, with open-source models gaining traction due to their cost-effectiveness and flexibility [13][16] - The focus of enterprise-level model application is shifting from seeking a single powerful model to finding optimal solutions tailored for specific business scenarios, emphasizing cost-performance ratio, system flexibility, and security [6][20] Summary by Sections Introduction - The report, published by Frost & Sullivan in collaboration with the Head Leopard Research Institute, surveys 700 IT department heads, technical directors/managers, and AI project leaders across various industries including finance, manufacturing, internet, consumer electronics, and automotive [4][28] - The study aims to assess the deployment of open-source and closed-source models in the enterprise-level GenAI market and to provide structured insights into the current application status and trends [4] Section 1: Overview of Enterprise-Level GenAI Development - The development of enterprise-level GenAI is characterized by the parallel growth of open-source and closed-source models, with open-source models becoming the preferred choice for low-cost implementation and autonomy [13][16] - Open-source models are increasingly recognized for their adaptability and long-term value, while closed-source models are favored for their reliability and performance [13][16] Section 2: Current Status and Trends of Model Invocation - The daily invocation of enterprise-level models has surged, indicating a shift from pilot testing to large-scale implementation, with significant implications for resource consumption and industry restructuring [18][19] - Key drivers of this growth include the expansion of model and computing power supply, accelerated deployment in various sectors, and the emergence of ecosystem effects that enhance efficiency [19][20] Section 3: Analysis of Model Invocation Behavior - The choice between open-source and closed-source models is primarily driven by business value, with open-source models offering greater flexibility and control, while closed-source models provide reliability and ease of use [24][26] - The top factors influencing the selection of open-source models include performance, customization ease, and knowledge ownership, whereas closed-source models are chosen for their reliability and brand reputation [25][26][27]