AD
MiniMax AI logo

MiniMax AI

Freemium🇨🇳DecacornHigh Burn

China's frontier multimodal AI — text, video, voice, music

78

Overall score

48

Heat score

Pricing

Free$0/month
Agent Basic$19/month
Agent Pro$69/month
API Token PlanUsage-based
Speech SubscriptionFrom $9.99/month

Technical Specs

Inputs

Text Prompt, Image, Audio File, Video Clip, Document, URL, Voice Sample

Outputs

Generated Text, AI Video, Text To Speech Audio, AI Music, AI Image, Code, Agent Task Output, Voice Clone

AI Type

Multimodal

Model Architecture

MoE Transformer

Daily Prompts

N/A

Context Length

4M

Output Quality

Accuracy

85%

Content

84%

Reasoning

83%

Company Profile

Company

MiniMax Group Inc.

Founded

2021

HQ

Shanghai, China

Employees

385

Total Raised / Total Funding

$1.15B

Revenue

$79M

Valuation

$12.8B

ARR

$94.8M

CEO

Yan Junjie

Overview

Estimated Paid Users

1.77M

Current estimate

Total Earnings Till Date

$112.5M

+3.41% from last month

Market Share

1.8%

Current share

Average Session

18

Per active user

Hallucination Rate

12%

Model quality signal

Growth Rate

+5.26%

Monthly active users

Burn Rate

$270M

Total expenses / years active

Paid User Gain

+22.73%

Monthly paid user trend

Profit Analysis

-$750M

Total Loss

$1.3B

Total Profit

$0

Performance Metrics

Accuracy

85%

Context

84%

Reasoning

83%

Safety

72%

Benchmarks

SWE-bench Verified

80.2%

SWE-bench Verified

51.3%

MiniMax AI Models

MiniMax-Text-01

Type: Text

Description: 456-billion-parameter mixture-of-experts LLM with Lightning Attention mechanism supporting a 4-million-token context window. Open-sourced on Hugging Face. Standard API: $0.20/M input, $1.10/M output tokens.

Context Length: 4M tokens

Architecture: MoE Transformer

MiniMax-VL-01

Type: Multimodal

Description: Multimodal vision-language model released alongside Text-01, capable of understanding and reasoning over images and text with a long context window. Open-sourced.

Context Length: 4M tokens

Architecture: MoE Transformer

MiniMax-M1

Type: Reasoning

Description: Reasoning-focused model with hybrid architecture requiring 70% fewer computational resources than comparable frontier models. Features 8x longer context than DeepSeek-R1 and excels at agentic long-horizon tasks.

Context Length: 4M tokens

Architecture: MoE Transformer

MiniMax-M2

Type: Text

Description: Agentic foundation model optimized for coding, tool use, and multi-step task execution with 10B activated parameters (230B total). Priced at $0.30/M input, $1.20/M output — ~8% of Claude Sonnet pricing.

Context Length: 4M tokens

Architecture: MoE Transformer

MiniMax-M2.5

Type: Text

Description: SOTA agentic model scoring 80.2% on SWE-Bench Verified and 51.3% on Multi-SWE-Bench. Runs at ~100 tokens/second. Priced at $0.20/M input, $1.17/M output. Designed for real-world productivity at minimal cost.

Context Length: 4M tokens

Architecture: MoE Transformer

Funding Rounds & Investors

Total Funding

$1.15B

Rounds

6

IPO

$618M

Jan 2026

Listed on Hong Kong Stock Exchange (HKEX: 02155); raised ~HKD 4.8B (~$618M); shares more than doubled on debut; source: VnExpress, Founded.com Jan 2026

Series D+

$300M

Jul 2025

Series B extension led by Shanghai state-owned capital at $4B valuation; source: Sacra, Reuters July 2025

Series B

$600M

Mar 2024

Led by Alibaba at $2.5B valuation; source: Fortune, Bloomberg, SiliconAngle Mar 2024

Founders/Team

YJ

Yan Junjie

Co-Founder, Chairman, CEO & CTO

YY

Yun Yeyi

Co-Founder & COO

ZY

Zhou Yucong

Co-Founder & Director

Direct competitors

No direct competitors available.

Change Log / Major Updates

2025 · Jun 1

MiniMax launches MiniMax-M1, a reasoning model requiring 70% fewer computational resources than comparable models, with 8x longer context than DeepSeek-R1. Competing on agentic and coding benchmarks.

2026 · Jan 9

MiniMax lists on HKEX at HKD 345/share, raising ~$618M. Shares more than double on opening day, achieving a market cap exceeding $12.8B and oversubscribed 1,838 times — the only HK tech IPO to double in four years.

2026 · Feb 1

MiniMax releases M2.5, scoring 80.2% on SWE-Bench Verified and 51.3% on Multi-SWE-Bench. M2.5-Lightning delivers 100 tokens/second at $0.30/M input tokens — positioned as intelligence too cheap to meter.

Compliance, Integrations & Support

Industry: Not specified

Compliances: Not specified

Integrations: Alibaba Cloud, Hugging Face, vLLM, SGLang, OpenRouter, Claude Code, Cursor, Cline, MiHoYo, Tencent, iOS App Store, Google Play, GitHub

Support:email, help center, community forum, enterprise support, api

Target audience: Developers, Content Creators, Enterprises, Game Studios, Mobile App Users, AI Researchers, Startups, Marketing Teams, Media Producers

Supported languages: English, Chinese, Spanish, French, German, Japanese, Korean, Portuguese, Arabic, Russian, Hindi, Italian, Dutch, Turkish, Polish, Vietnamese, Thai, Indonesian, Malay, Swedish

MiniMax AI Acquisitions

No acquisition records available.

AD

Reviews & Rating

0 reviews

No reviews yet

Be the first to share how MiniMax AI performs for your workflow.

0.0

Accuracy

0.0

Ease of Use

0.0

Output Quality

0.0

Security

0.0

More About MiniMax AI

In December 2021, while SenseTime was preparing its landmark Hong Kong IPO, one of its youngest vice presidents quietly handed in his resignation. Yan Junjie — mathematician, AI engineer, avid gamer — had watched a bot defeat world-class Dota 2 players in 2019 and became consumed by a single question: could general intelligence be engineered, methodically, like a product? The company he founded in the final days of 2021, a full year before ChatGPT launched, is now called MiniMax.

What MiniMax Actually Builds

MiniMax is one of the few AI companies anywhere that has built frontier models across every major modality from scratch: text (the MiniMax-Text-01 at 456 billion parameters with a 4-million-token context window), video (Hailuo-02 generating 1080p clips with physics-consistent camera controls), speech (Speech-02 supporting 40+ languages with sub-second latency), and music (Music-01 composing 4-minute songs with vocals). This vertical integration is not an accident — it is the strategic bet that quality control and margin improvement require owning the full stack. By end of 2025, MiniMax's speech model had generated over 200 million hours of audio globally; its video models had produced 600 million clips.

Products and Traction

Consumer-facing, the company runs Talkie — an AI companion app that by mid-2024 ranked as the fourth most-downloaded AI app in the United States, surpassing Character.AI, with 11 million monthly active users. By December 2024, Talkie's MAU had grown to nearly 30 million. The Hailuo AI video platform similarly accelerated after its September 2024 launch. On the enterprise side, MiniMax's Open Platform serves 214,000 developers and enterprise customers across 100+ countries, offering API access at pricing it positions as a fraction of comparable Western models — the M2.5 API costs approximately 8% of Claude Sonnet's equivalent rate.

The company listed on the Hong Kong Stock Exchange on January 9, 2026, with shares more than doubling on debut day — the only Hong Kong tech IPO in four years to achieve that milestone — and oversubscribed 1,838 times. Full-year 2025 revenue reached $79 million, more than doubling from $30.5 million in 2024, driven by Hailuo AI monetization and enterprise API adoption. Gross margin improved from 12.2% in 2024 to 25.4% in 2025, reflecting model efficiency gains. As of December 2025, MiniMax had served 236 million cumulative users across 200+ countries.

MiniMax AI FAQ's

What is MiniMax AI?

MiniMax is a Shanghai-based AI company that builds full-modality foundation models covering text, video, speech, music, and image generation. It operates consumer apps including Talkie (AI companion), Hailuo AI (video generation), and Xingye, as well as an enterprise Open Platform with API access for developers. MiniMax listed on the Hong Kong Stock Exchange in January 2026.

What models does MiniMax offer?

MiniMax's flagship models include MiniMax-Text-01 (456B parameter LLM with a 4-million-token context window), the M2-series agentic models (M2, M2.1, M2.5, M2.5-Lightning) optimized for coding and tool use, Hailuo-02 (1080p video generation), Speech-02 (TTS in 40+ languages), and Music-01 (4-minute song composition). Most models are available via API on platform.minimax.io.

How much does the MiniMax API cost?

MiniMax-M2.5 is priced at $0.20 per million input tokens and $1.17 per million output tokens. MiniMax-M2.5-Lightning costs $0.30 per million input tokens and $2.40 per million output tokens. MiniMax-Text-01 is priced at $0.20 per million input tokens and $1.10 per million output tokens. Video generation starts at $0.10 per 6-second clip. These rates are among the lowest in the industry for frontier-class models.

Is MiniMax open source?

Partially. MiniMax has open-sourced the weights for MiniMax-01 (Text-01 and VL-01) and the M2 and M2.1 model families on Hugging Face. The weights can be deployed locally using vLLM or SGLang. Commercial models like M2.5 are available via API but not fully open-sourced.

What is Hailuo AI?

Hailuo AI is MiniMax's text-to-video generation platform, launched in March 2024 with the video-01 model. Users enter text prompts to generate 6–10 second, 1080p video clips. As of end of 2025, MiniMax's video models had helped creators generate over 600 million videos worldwide. Hailuo AI is available as a web app and mobile application.