Analysis of Business Model Differences Between Zhipu AI and MiniMax

Based on the collected information, I will systematically analyze the business model differences between Zhipu AI and MiniMax.

Zhipu AI’s business model centers on

. According to its prospectus data, the MaaS business contributed 84.8% of the company’s revenue in the first half of 2025 [1]. Through its open platform bigmodel.cn, the company provides model API calling services to developers and enterprises, having served over 400,000 enterprise users and more than 8,000 institutional clients across over 10 industries including finance, education, internet, retail, automotive, energy, and traditional manufacturing [2].

MiniMax adopts a completely different strategy, focusing on

. As of September 30, 2025, its C-end revenue accounts for over 71%, while B-end revenue accounts for only 28.9% [3]. The company has built a complete product matrix, including applications such as Talkie/Xingye, Hailuo AI, MiniMax App, and MiniMax Voice, with cumulative users exceeding 212 million and monthly active users (MAU) reaching 27.622 million [4].

Dimension	Zhipu AI	MiniMax
Core Revenue Source	B-end MaaS/API Calls (84.8%)	C-end AI Native Products (71.1%)
2024 Revenue	Approximately $42 million	$30.52 million
2025 First Three Quarters Revenue	Not Disclosed	$53.44 million (175% YoY Growth)
Revenue Growth Driver	Developer Platform, Enterprise Clients	Talkie/Xingye, Hailuo AI
Gross Margin	Not Disclosed	23.3% (First Nine Months of 2025)

Zhipu AI’s commercialization path is closer to the traditional models of OpenAI and Anthropic, allowing developers to call at low cost by providing general intelligent capabilities, quickly achieving large-scale model call volume [5]. MiniMax is more like an ‘AI product factory’, and founder Yan Junjie once said: ‘In the large model era, the real product is actually the model itself’ [6].

Zhipu AI’s product strategy focuses on

:

• MaaS Platform
  : Provides model matrix access, agent development tools, model fine-tuning platform
• Deployment Solutions
  : Offers cloud privatization, local privatization, and hardware-software integrated solutions based on different customer needs
• C-end Products
  : Zhipu Qingyan (approximately 25 million users), but its revenue contribution is relatively small
• Technology Route
  : Full-stack independent research and development, GLM series large models target GPT-4

MiniMax chooses the

strategy:

• Core Products
  : Talkie/Xingye (virtual social platform), Hailuo AI (video generation), MiniMax Voice
• Technical Advantages
  : Full-modal coverage (text, voice, video), leading performance of M1/M2 models
• Market Strategy
  : Focuses on content tools, content platforms, agents, etc., and does not participate in traditional dialogue product competition
• Overseas Layout
  : Overseas revenue accounts for over 73%, with Singapore and the US contributing 24.3% and 20.4% of revenue respectively [7]

Zhipu AI adopts a highly competitive pricing strategy. CEO Zhang Peng once said: ‘Our models are good enough—among the world’s top echelon. We have huge advantages in price and cost.’ The company even joked that if Anthropic’s product sells for $200, Zhipu AI will sell for 200 RMB [8]. This strategy has enabled the company’s API business to grow rapidly, with daily service volume exceeding 60 billion tokens.

MiniMax’s revenue mainly comes from:

• Subscription Revenue
  : 1.39 million paying users for Talkie/Xingye, with an average expenditure of $5 per user
• In-App Recharge
  : 310,000 paying users for Hailuo AI, with an average expenditure of $56 per user
• Enterprise Services
  : Average expenditure of $6,167 per paying customer on the open platform
• Value-Added Services
  : In-app recharge revenue of Hailuo AI accounts for 6.2%, user subscriptions account for 26.4% [9]

Indicator	Zhipu AI	MiniMax
Technical Architecture	Independent R&D of GLM Series	Mixture of Experts (MoE) Architecture
Model Capability	GLM-4.5 Ranks Top 3 Globally	Leading Performance of M1/M2 Models
R&D Personnel Ratio	Not Disclosed	73.77% (284/385 Employees)
R&D Investment	Approximately 70% of Raised Funds for R&D	$180 million in First Nine Months of 2025
Full-Modal Capability	Text-to-Image (CogView4), Video Generation (CogVideoX)	Voice (Speech-02), Video (Hailuo-02), Text (M2)

Both companies take R&D as core investment, but MiniMax has higher per capita output efficiency: among 385 employees, R&D personnel account for 73.77%, with an average age of only 29 [10]. Zhipu AI relies on Tsinghua University’s academic background, adheres to the ‘independent innovation’ route, and GLM series models have been adapted to more than 40 domestic chips.

Zhipu AI submitted listing guidance filing to the Beijing Securities Regulatory Bureau of the China Securities Regulatory Commission on April 14, 2025, and is expected to list on the A-share market as early as 2026 with a valuation exceeding $3 billion (approximately 24.4 billion RMB) [11]. The company plans to use approximately 70% of the raised funds to continuously enhance general AI large model R&D capabilities.

MiniMax first published the post-hearing data set on December 21, 2025, which is expected to set a new record for the shortest time from establishment to IPO. Founded in early 2022, the company completed its IPO sprint in only about 4 years, with well-known investors such as Tencent and Alibaba participating in multiple rounds of financing [12].

Dimension	Zhipu AI	MiniMax
Business Model	B-end Technology Output (MaaS)	C-end Product Driven (AI Native Applications)
Revenue Structure	84.8% from MaaS	71.1% from C-end Products
Market Focus	Mainly Chinese Market (covering 8,000+ institutions)	Global Layout (73% revenue from overseas)
Technology Strategy	Full-stack Independent R&D	Model as Product
Competitive Advantage	Low-price Strategy, Government Relations	Product Innovation, Overseas Market
Profit Path	Scaling of Developer Ecosystem	Monetization of Product Matrix

• More stable B-end revenue base
• Policy support from the government for independent and controllable AI technology
• Lower customer acquisition cost (platform effect)
• Academic resource support from Tsinghua University

• Faster revenue growth rate (175% YoY)
• Wider global user base
• Higher gross margin improvement potential (from -24.7% to 23.3%)
• Younger and more innovative team

• Zhipu AI: B-end customer concentration, price competition pressure
• MiniMax: C-end user retention, continuous losses (net loss of $512 million in the first nine months of 2025)

Overall, Zhipu AI and MiniMax represent two distinct commercialization paths for AI large model companies: the former is a technology platform type, achieving scale through developer ecosystems and enterprise clients; the latter is an application type, directly reaching consumers by creating popular applications. Both models have their own advantages and disadvantages, and their future development paths are worth continuing to pay attention to.

---

[1] 36Kr - Going Overseas, Listing: China’s First Batch of Large Models Have Made It
[2] Ifanr - Dialogue with Zhipu AI CEO Zhang Peng
[3] Securities Times - MiniMax’s Lightning Rush to Hong Kong Stocks: C-end Revenue Exceeds 70%
[4] Sina Finance - Xiyu Technology Strives for the First Global Large Model Stock
[5] Wall Street CN - MiniMax Discloses Performance for the First Time
[6] 21st Century Business Herald - MiniMax’s Lightning Rush to Hong Kong Stocks
[7] Sina Finance - MiniMax’s Lightning Rush to Hong Kong Stocks: C-end Revenue Exceeds 70%
[8] Wall Street CN - Targeting OpenAI! Zhipu AI Vies for “First Domestic AI Stock”
[9] Nanfang + - Competing for “First Global Large Model Stock”
[10] OFweek - Hong Kong IPO Zhipu Huazhang
[11] 36Kr - Zhipu Secures $300 Million Financing
[12] 36Kr - Dissecting Zhipu