Skip to main content

MiniMax

Category: Foundation Models / LLMs

A leading generative AI company providing multimodal foundation models and consumer-facing applications for social interaction, productivity, and media generation. MiniMax was founded in 2021. The company is led by Yan Junjie. Based in Shanghai, China. Team size: 101-500. Total funding raised: $1.77B. Latest round: Venture Round ($300.0M, Jul 2025). Key investors include Alibaba, Tencent, MiHoYo, HongShan, Hillhouse Investment, IDG Capital.

AMW Analysis

MiniMax develops multimodal foundation models and consumer applications for social interaction, productivity, and media generation. Its recent product activity centers on M3, an open-weight model combining one-million-token context, native text-and-image capabilities, and coding-agent features. The company has positioned M3 around long-context efficiency through sparse attention and Token Plan pricing, while its open-source release included Day-0 support for Moore Threads’ MTT S5000 GPU. This indicates an effort to pair model distribution with domestic hardware integration.

The news flow also shows continued expansion across creative and agent-oriented products. Music 2.6 added faster initial generation, AI-generated covers, broader instrument support, and a Music Skill for AI-agent integration, while MaxClaw contributed to the company’s agent product presence. Financially, MiniMax completed a 16 billion HKD funding round in July 2026 following its July 2025 venture round, and has also prepared for a mainland China offering after its Hong Kong listing.

AMW analysis, generated from 8 tracked news signals.

Founded
2021
Headquarters
Shanghai, China
Team size
101-500
Total funding
$1.77B

Value proposition

Combines high-emotional-intelligence conversational AI with professional-grade, cost-effective multimodal generation powered by proprietary efficient architectures.

Products and solutions

Talkie (AI Character Social App), Hailuo AI / Hailuo Video (Text-to-Video Generation), MiniMax-M1 / abab Series (Foundation Models), Music-01 (AI Music Generation), Speech-02 (High-fidelity TTS)

Unique value

Successfully bridges deep model research with viral consumer apps (Talkie); pioneered 'Lightning Attention' for massive context windows (4M+ tokens) at low cost.

Target customer

Gen Z consumers (B2C), Content Creators, Enterprise Developers (B2B)

Industries served

Social Media & Entertainment, Content Creation, Enterprise Software, Gaming

Technology advantage

Proprietary Hybrid MoE (Mixture of Experts) + Linear Attention architecture delivers superior inference speed and long-context handling compared to traditional Transformers.

How they differentiate

Dual-engine strategy combining viral consumer apps (Talkie) with proprietary 'Lightning Attention' architecture for efficient long-context processing.

Main competitors

Moonshot AI (Kimi), Character.ai, Zhipu AI

Key partnerships

Alibaba, Tencent, MiHoYo, Volcano Engine

Notable customers

Kingsoft Office (WPS), Xiaoice, Millions of Talkie B2C Users

Major milestones

Listed on HKEX on Jan 9, 2026 (stock code 0100), raised $619M, Talkie app topped US App Store Entertainment charts in 2024, Launched Hailuo Video (S2V-01) rivaling global SOTA models, Released MiniMax-M3 in June 2026 with frontier coding & 1M context

Growth metrics

1.8M+ paying users (2025); Revenue $79M in 2025 (+159% YoY); ARR exceeded $150M by Feb 2026; 70%+ revenue from overseas.

Market positioning

Leader in Consumer GenAI (China's 'AI Tiger') with strong global B2C footprint.

Geographic focus

Global (North America/SEA for Consumer), China (Enterprise/B2B)

Patents and IP

Core IP revolves around 'Lightning Attention' mechanism and efficient MoE scaling; multiple patents filed for conversational logic and multimodal processing.

About Yan Junjie

Former Vice President at SenseTime; led the deep learning toolchain and underlying algorithm teams; PhD from Institute of Automation, Chinese Academy of Sciences (CASIA).

Latest news about MiniMax

More Foundation Models / LLMs companies

Official website: