MiniMax
MiniMax is a Chinese AI foundation model company founded in 2021 by former SenseTime executives. It has emerged as one of China’s “AI Tigers” alongside DeepSeek, Baichuan, Zhipu, and Moonshot.
Key Facts
- Headquarters: China
- Founded: 2021
- Valuation: ~4 billion IPO (2025)
- Funding: $850 million VC (backed by Alibaba and Tencent)
- Users: 212 million individuals across 200+ countries, 100,000+ enterprises
- Revenue: $53.44 million (first 9 months 2025, +175% YoY)
- IPO Filing: December 2025 with Hong Kong Stock Exchange
Model Series
Foundation Models
- MiniMax-Text-01 - Natural language processing model with 1M token context
- MiniMax-VL-01 - Multimodal vision-language model
- MiniMax M1 - Open-weight foundation model (June 2025)
- MiniMax M2.1 - Coding-focused MoE model, 230B params (December 2025)
Audio Models
- MiniMax Speech-02 - Text-to-speech model, 30+ languages, voice cloning
- T2A-01-HD - Speech generation model
Video Models
- Hailuo AI Video - Text/image to video generation, ranked #2 globally
Products and Platforms
Consumer Applications
- Hailuo - Flagship generative AI platform
- Talkie - Social AI application
- Glow - AI companion app with customizable virtual characters (launched Oct 2022)
- Xingye - AI companion community for Chinese users (5.25M MAU Nov 2024)
Enterprise Tools
- MiniMax Agent - AI-native agent application for enterprise workflows
- MiniMax Audio - Audio generation capabilities
- Open API Platform - Developer and enterprise integration tools
Competitive Position
Competes with OpenAI GPT-4o, Google Gemini 1.5 Pro, and Meta LLaMA series. Known for long-context processing (up to 1 million tokens) and multimodal capabilities spanning text, image, audio, and video.
Technology Highlights
- Hybrid Mixture-of-Experts architecture
- Proprietary Lightning Attention mechanism
- 75% reduction in compute vs comparable models (M1)
- CISPO training method halving training steps needed