MiniMax
Chinese AI company building multimodal foundation models for text, audio, image, video, and music.
Updated April 2026
Overview
- Website
- minimaxi.com
- Founded
- 2021
- Headquarters
- Shanghai, China
- Segment
- Frontier Foundation Model Labs
- Posture
- Regional / Emerging
Product overview
MiniMax develops proprietary multimodal LLMs like MiniMax-M2.7 series (text/agentic), Hailuo video generation, Speech-02 TTS/STT, image-01, and music models, accessible via API platform. Consumer apps include Talkie (AI characters, 11M MAU), Hailuo AI (text-to-video), and MiniMax Agent. Used by over 200M global users and 214K enterprise clients/developers, standing out with ultra-long context (1M tokens), efficient MoE architecture, and strong Asian language support versus Western labs.
Revenue model
API pay-per-use ($0.3/M input/$1.2/M output tokens for M2.7; $0.19 per 6s video) and monthly subscriptions (Starter $10/mo, Max $50/mo; up to $150/mo highspeed); consumer app in-app purchases and AI-native product revenue ($53M in 2025).
Moat
MiniMax's competitive moat stems from its native multimodality architecture trained simultaneously across text, speech, music, and video from inception, combined with a diversified consumer-to-enterprise revenue model that generates rapid iteration loops and viral adoption optionality. This integrated multimodal foundation—evidenced by its 88.6% VIBE score—creates genuine technical differentiation difficult to replicate, while its consumer products (Talkie at 35% of revenue, Hailuo AI at 33%) serve as both profit centers and R&D laboratories that accelerate model improvement at lower cost than pure enterprise competitors.