MiniMax
MiniMax is a general AI technology company founded with the mission to co-create intelligence with everyone, dedicated to achieving Artificial General Intelligence (AGI) through multimodal foundation models capable of understanding and generating text, audio, images, video, and music.
At a Glance
- Enterprise software developers and AI agent creators
- Digital content creators and influencers
- Short video and podcast producers
- Online literature and audiobook publishers
- +12 more
AI Tools by MiniMax
(2)
MiniMax - Your AI Agent
Mobile AI Assistant App iOS
MiniMax Agent
Multimodal AI Agent Platform
Discussions
No discussions yet
Be the first to start a discussion about MiniMax
Latest News
MiniMax launches M2.5 and M2.5-Lightning models for real-world productivity, priced at $1 per hour
Forge: Scalable Agent RL Framework and Algorithm released
MiniMax to Report 2025 Full Year Financial Results on March 2, 2026
Hailuo AI x fal partnership announced in Istanbul: A New Era for Creativity
Products & Services
Frontier language models optimized for coding, agentic tool use, search, and office productivity. M2.5 achieves 80.2% on SWE-Bench Verified. M2.5-Lightning offers 100 tokens/second throughput.
Model specialized for multi-character roleplay and long-horizon interaction scenarios.
Enhanced model for complex tasks with improved multilingual and code engineering capabilities, supporting Rust, Java, Golang, C++, Kotlin, and JavaScript.
Open-weight reasoning model with 230B parameters (10B active) using MoE architecture, designed for coding and agentic workflows.
Market Position
MiniMax positions itself as a cost-effective, high-performance alternative to Western AI giants (OpenAI, Anthropic, Google) with emphasis on open-source strategy and multimodal capabilities. The company claims its M2.5 model operates at 8% the cost of Claude 3.5 Sonnet while achieving competitive benchmark performance. Its M1 model uses Apache licensing for unrestricted commercial use, contrasting with more restrictive licenses from Meta and DeepSeek. MiniMax differentiates through: (1) Lightning Attention architecture requiring 70% less computational power than DeepSeek R1; (2) 1 million token context window matching Gemini 2.5 Pro; (3) Vertical integration with consumer apps (Talkie: 11M MAU) and enterprise APIs; (4) Strong Chinese market position backed by Alibaba and Tencent; (5) Focus on real-world productivity and agent-native design. The company competes with domestic rivals (DeepSeek, Zhipu, Moonshot AI, Baidu Ernie, Alibaba Qwen) and positions itself as leading the open-source race in China while targeting the global market, particularly the US, Philippines, UK, and Canada.
Leadership
Founders
Yan Junjie
PhD from the Institute of Automation, Chinese Academy of Sciences. Former Vice President and Executive Dean of the Research Institute at SenseTime for over six years, where he managed 700 engineers. Expert in smart cities, gaming, and facial recognition. Left SenseTime in December 2021 to found MiniMax.
Yeyi Yun (Yun Yeyi)
Graduate of Johns Hopkins University and Columbia University. Former Head of Strategy in the CEO office and Director of Innovative Business at SenseTime. Spent four years in the United States and brings operational discipline and global orientation to the company.
Yang Bin
Former computer vision researcher at SenseTime. Co-founded MiniMax in December 2021.
Zhou Yucong
Former computer vision researcher at SenseTime. Co-founded MiniMax in December 2021.
Executive Team
Yan Junjie
Chairman and CEO
36 years old (as of 2026). PhD from Institute of Automation, Chinese Academy of Sciences. Former Vice President at SenseTime for 6+ years, managing 700 engineers. Founded MiniMax in December 2021.
Yeyi Yun (Yun Yeyi)
Co-Founder and Chief Operating Officer
Graduate of Johns Hopkins University and Columbia University. Former Head of Strategy in CEO office and Director of Innovative Business at SenseTime. Served as director and COO since December 2022, re-designated as Executive Director in June 2025.
Board of Directors
Founding Story
MiniMax was founded in December 2021 by Yan Junjie and three other former SenseTime executives. Yan left SenseTime just before its IPO, motivated by a desire to build AI models that could compete with OpenAI and challenge US dominance in AI. He had been tracking OpenAI since 2019, particularly inspired by their Dota 2 achievements. When he initially pitched his vision of building multi-modal AI to China's largest tech companies and investors, many believed it was a fraud. A critical early backer was gaming giant miHoYo. The company was established to build AI capable of processing text, images, audio, and video together, with a focus on making video and audio production faster and cheaper for creators and businesses.
Business Model
Revenue Model
MiniMax operates a dual-sided business model combining consumer applications and enterprise API services. Revenue comes from: (1) Consumer subscriptions for AI-native apps (Talkie, Hailuo AI) with freemium models at $9.99/month for premium; (2) API usage fees for enterprise clients on pay-as-you-go basis; (3) Pre-purchased resource packages for video and audio generation; (4) Subscription plans for developers (Coding Plan). The company practices vertical integration by training proprietary models. Revenue is balanced across three segments (1:1:1): companion AI apps, content generation tools, and API services.
Pricing Tiers
Premium subscription for Talkie AI companion app with ad-free experience
100k credits, 10 voice slots, 10 RPM
300k credits, 100 voice slots, 50 RPM
1.1M credits, 250 voice slots, 200 RPM
3.3M credits, 500 voice slots, 500 RPM
20M credits, 800 voice slots, 800 RPM
3,760 units, 20 RPM, 5% savings
9,920 units, 30 RPM, 10% savings
18,900 units, 40 RPM, 15% savings
26,780 units, 50 RPM, 20% savings
Input/Output tokens. Approximately $1 per hour at 100 tokens/second
Input/Output tokens. Approximately $0.30 per hour at 50 tokens/second
Input/Output with prompt caching at $0.03 read/$0.375 write per million tokens
speech-2.6-turbo and speech-02-turbo models
speech-2.6-hd and speech-02-hd high-definition models
768P 6s ($0.19), 768P 10s ($0.32), 1080P 6s ($0.33)
512P from $0.10, 768P from $0.28, 1080P from $0.49
Up to 5 minutes per music track
image-01 model generation
Rapid voice cloning and voice design
Target Markets
- Enterprise software developers and AI agent creators
- Digital content creators and influencers
- Short video and podcast producers
- Online literature and audiobook publishers
- Film and media production companies
- Call centers and customer service operations
- Software development and code engineering
- Professional office automation and productivity
- Creative content generation (video, music, audio)
- Multi-character roleplay and social interaction
- Immersive AI companions and virtual characters
- Audiobook production and online literature
- miHoYo
- Haivivi
- Yuewen
- Guagua Audiobooks