一文读懂Minimax招股说明书：领先的通用多模态大模型平台，AI原生应用矩阵+开放式生态驱动商业化落地

1. Report Industry Investment Rating No relevant content provided. 2. Core Views of the Report - The company is one of the core providers of general multi-modal large models and entered the stage of large-scale commercialization in 2025. It is positioned as a provider of general multi-modal large models and AI-native applications, with deep technical accumulation in voice generation, multi-round dialogue, and multi-modal interaction. It is in the first echelon among domestic general large model manufacturers [3]. - The company's business model is centered around self-developed general large models, and its revenue growth is continuously driven by the increasing volume of model calls. In 2025, its revenue continued the high-growth trend. The company has comprehensive competitive advantages, including continuous iteration of general multi-modal model capabilities, parallel product and commercialization paths for B and C ends, a platform-based and scalable business model, and a management and R & D team with long-term experience in the AI field [4][5]. 3. Summary According to the Table of Contents 3.1 Company Overview 3.1.1 Growth Review - The company was founded in 2021, focusing on general artificial intelligence and the research and development of self-developed large language models and multi-modal models. It has gradually built a relatively complete general multi-modal model system and application matrix. As of before the IPO, it had completed about 7 rounds of financing, with a cumulative financing amount exceeding $1.5 billion. The actual controller of the company is Dr. Yanjunjie, and Alibaba is the largest external institutional shareholder [13][15][18]. 3.1.2 Main Business - AI-native products (mainly ToC): The company has launched a number of AI-native applications for individual users, including MiniMax (intelligent Agent), Hailuo AI (multi-modal content creation), MiniMax Voice (voice synthesis and interaction), and Talkie/Xingye (AI character companionship and interaction), aiming to achieve commercialization through subscriptions, virtual content consumption, and the spillover of capabilities in the medium to long term [21][22][23]. - Open platform and other AI enterprise services (ToB/developers): The company provides services such as model ability opening API (MaaS), enterprise-specific reasoning resource pools, model authorization and deployment, and cross-industry enterprise solution support to enterprises and developers, with various charging methods [28]. - Business model - MaaS: The company provides self-developed general large model capabilities to external customers in the form of cloud services. The user scale and core operating indicators are driven by AI-native products (ToC) and the open platform (ToB) [33]. - Pricing strategy: The company adopts a multi-dimensional and hierarchical pricing strategy. The ToC end mainly uses a monthly subscription system, supplemented by prepaid points or virtual item recharge; the ToB end uses API packages and token-based pay-as-you-go billing [35]. - Customer structure: The company's customers are diverse and international, with enterprise customers as the core revenue source. The customer concentration has been continuously decreasing, and the company uses multiple channels to acquire customers and signs framework agreements to ensure long-term stable customer relationships [41]. 3.1.3 Financial Analysis - Revenue: The company's revenue has grown rapidly since the start of commercialization. AI-native products (ToC) contribute the current main revenue scale, and the open platform and enterprise services (ToB) are growing rapidly. Overseas revenue accounts for a high proportion [48]. - Gross profit and expense ratio: The company's overall gross profit margin has been significantly repaired, and the expense ratio has been rapidly converging. The company maintains high R & D investment to support long-term competitiveness [55]. 3.2 Industry Overview 3.2.1 Technological Evolution Trends of Large Models - Scaling Law: The focus has shifted from simply expanding scale to improving training efficiency and generalization ability under controllable costs and stability [62]. - Cost reduction: The unit cost of intelligence has been continuously decreasing through model structure optimization, reasoning acceleration, and computing power scheduling improvement [62]. - Agent application: Large models have evolved from single-point generation to Agents with task decomposition, tool invocation, and multi-step execution capabilities [62]. - Multi-modal: The multi-modal capabilities of text, voice, image, and video are accelerating integration, moving from multi-model splicing to unified modeling [62]. 3.2.2 Changes in the Market Pattern of Large Model Applications - The application market of large models shows a hierarchical structure, with different levels having different representative products, target users, core capabilities, commercialization models, and competition points [63][64]. 3.2.3 Market Size and Competition Pattern of Large Models - Market size: The global large model market is in a stage of rapid growth, with the market size expected to increase from about $10.7 billion in 2024 to about $206.5 billion in 2029, with a CAGR of 80.7% [68]. - Market structure: The large model-related revenue is divided into MaaS and application, with the application layer having a higher growth rate and becoming the core engine of market expansion [69]. - Competition pattern: The large model industry chain shows a hierarchical competition structure, with the basic model and MaaS layer dominated by a few leading manufacturers, and the application and Agent layer showing a diversified competition situation [69]. 3.3 Core Competitiveness 3.3.1 Long - term Barriers Built by a Full - Modal Unified Base and Engineering Efficiency Advantages - Full-modal capabilities: The company uses a full-modal integrated approach, which can output consistent and scalable intelligent capabilities in various scenarios, reducing cross-modal development and integration costs [71]. - Model algorithm innovation: The company focuses on performance, cost, and deployability, using architectures such as MoE, linear attention mechanisms, and CISPO reinforcement learning algorithms [71]. - Cost advantage: The company has the ability to systematically reduce costs in training and reasoning, providing more competitive pricing strategies and broader customer coverage [71]. 3.3.2 Strategy and Commercialization: Scalable Architecture + Dual - Wheel Drive of ToC/ToB to Amplify Scale Effects - Scalable architecture: The company uses a highly modular and horizontally scalable system architecture, which can maintain system stability and delivery quality with the growth of token calls and the expansion of the customer base [72]. - Commercialization path: The company adopts a parallel strategy of AI-native products and MaaS, balancing growth elasticity and revenue certainty [72]. - Open platform and customer stickiness: The company's open platform has become a key hub in its business model, with customers deeply embedding model capabilities into their products, increasing migration costs and forming a technology lock - in effect [72]. 3.4 Historical Financial Situation - Consolidated income statement: The report shows the company's income, cost, gross profit, and other items from 2022 to 2025, reflecting the company's operating performance and profitability changes [74]. - Consolidated balance sheet: It presents the company's assets, liabilities, and equity at different times, reflecting the company's financial position [75]. - Consolidated cash flow statement: It shows the company's cash inflows and outflows from operating, investing, and financing activities, reflecting the company's cash generation and utilization capabilities [78][79]. - Profit statement breakdown: It details the company's revenue and cost composition, including AI-native products and open platform and other AI-based enterprise services [80][81].