Business Segments
Business Segment Overview
MiniMax is a global AI foundation model company committed to developing multi-modal models from day one. The company operates across two key areas: Foundation Models — including MiniMax M Series (LLM), Hailuo-02 (video), Speech-02 and Music-02 (audio); and AI-Native Products — MiniMax Agent, Hailuo AI platform, MiniMax Audio, Talkie/Xingye entertainment platform, and the MiniMax Open Platform serving enterprise customers across 100+ countries. MiniMax adopted MoE architecture and hybrid attention early, enabling cost-efficient AI advancement with globally recognized performance.
Full-Stack AI Platform
Multimodal Capability: One of few AI labs with text, image, video, and audio under one roof.
Consumer + Enterprise: Dual revenue streams from B2C products and B2B APIs.
Video Differentiation: Hailuo AI has achieved genuine product-market fit globally.
China Optimization: Models specifically tuned for Chinese language and culture.
Foundation Model & Product Timeline
| 2022 | 2023 | 2024 | 2025 | |
|---|---|---|---|---|
| Large Language Model | abab 1 | abab 5.5 abab 6.0 (MoE) | Text-01 | M1 M2 |
| Video Generation | — | — | Hailuo-01 | Hailuo-02 |
| Audio Model | — | Speech-01 | Music-01 | Speech-02 Music-02 |
| Product Launches | — | Open Platform Talkie/Xingye | Hailuo AI | MiniMax Agent MiniMax Audio |
🧠 Foundation Models
📱 AI-Native Products
Model Suite Overview
Multimodal: Comprehensive models covering text, video, speech, and music generation.
R&D Excellence: Leveraged in-house research capabilities to maintain competitiveness across modalities.
Long Context: Models support extended context windows up to 1 million tokens.
Global Benchmarks: Competitive results on international AI benchmarks.
MiniMax has leveraged its R&D capabilities to build a comprehensive suite of foundation models, maintaining competitiveness across various modalities. The foundation model suite includes large language models, video generation models, and models for speech and music generation.
The flagship family of large language models comprising MiniMax-M1 and MiniMax-M2.
Open-source, large-scale hybrid-attention reasoning model. Adopts hybrid MoE architecture with lightning attention, enabling 1 million token context window and supporting AI agent development.
Engineered for elite performance in coding and agentic tasks. Data-efficient MoE architecture delivers higher-performance at faster inference speeds with optimized cost-efficiency.
Generates high-quality video content from a variety of information inputs.
Commercialized at scale with competitive results on global benchmarks upon release. Offers cinematic video quality, advanced prompt adherence, smooth motion, and style diversity. User-friendly interface with aesthetic refinement helps content creators and advertisers produce compelling videos from simple prompts.
Designed to generate natural, high-quality speech from text input.
Widely recognized as a top-performing speech model globally upon release in April 2025. Delivers hyper-realistic, personalized voice synthesis across multiple languages. Natural prosody and emotion control capabilities.
Key Technical Advantages
Product Suite Overview
Consumer Focus: Products designed for individual users with subscription-based access.
Global Reach: Products available in 100+ countries through web and mobile.
Foundation Model Powered: All products leverage MiniMax's proprietary models.
Organic Growth: Strong user adoption driven by product quality.
Leveraging the multi-modal foundation model suite, MiniMax delivers AI-native products and services that unleash the power of AI to benefit individual users, developers, and enterprise customers around the world. The evolution of these products is rooted in advancements in underlying foundation models — through continuous iterations and upgrades, MiniMax creates products with enhanced productivity and user experience.
MiniMax's intelligent AI agent application, designed to autonomously perform a wide range of tasks through natural language instructions. Supported by foundation models, MiniMax Agent can plan, reason, and execute complex actions such as coding, research, document drafting, and presentation creation within a unified workspace.
Fully integrates the Hailuo-02 model and has quickly become one of the world's most popular AI image and video creation platforms through organic user adoption. Offered in both web and app forms, designed for real-time, high-quality image and video generation.
Designed to provide users with high-fidelity audio generation capabilities. Accessible via web platform, MiniMax Audio integrates the Speech-02 model to support interactive audio synthesis and generate natural, high-quality speech from text input.
A globally recognized AI-native multi-modal entertainment platform. Users can engage with emotionally responsive AI themes or virtual characters powered by MiniMax's proprietary AI models. Talkie (international markets) / Xingye (Chinese domestic market).
Enterprise and developer-facing platform providing API access to MiniMax's foundation models. Serves customers across 100+ countries with text, speech, video, and image generation APIs, plus custom inference pools and licensed model deployments.
Revenue Streams
Dual Revenue Model: AI-native products (B2C) + Open Platform enterprise services (B2B).
Subscriptions: Premium access to MiniMax, MiniMax Audio, Hailuo AI, Talkie/Xingye.
Usage-Based APIs: Token-based billing for enterprise customers and developers.
Enterprise Services: Custom inference pools, licensed model deployments.
Primary Revenue Sources
Revenue from individual users through subscription-based access to consumer applications.
Premium access to MiniMax, MiniMax Audio, Hailuo AI, and Talkie/Xingye. Revenue recognized ratably over subscription period.
Users pre-purchase credits to recharge accounts. Consumable items recognized when used; non-consumable over estimated usage period.
Marketing services to mediation platforms. Revenue recognized at point-in-time when users view/click advertisements.
Revenue from enterprise customers and developers via usage-based APIs and custom services.
Token-based billing when customers call APIs. Recognized at point-in-time under agreed fee schedules.
Dedicated inference resources tailored to enterprise needs, ensuring stable and predictable model performance.
Foundation model licensing for customers to deploy in their own systems. Revenue recognized when control transfers.
Enterprise Service Verticals
Key Insights — What This Means
Unlike competitors focused on one modality (text or video), MiniMax builds ALL modalities in-house. This creates data flywheel effects and cost efficiencies that single-modality players can't match.
B2C products (Talkie, Hailuo AI) provide predictable subscription revenue, while B2B APIs offer high-margin enterprise deals. This diversification reduces dependency on any single revenue stream.
Hailuo AI achieved global adoption organically — a rare feat in AI. Video generation is where MiniMax truly stands out vs. text-focused competitors like OpenAI and Anthropic.
Early adoption of MoE (Mixture of Experts) architecture means MiniMax can train and run models at lower cost than dense-model competitors — critical given GPU access constraints in China.