Zhipu has initiated Pre-Deal Investor Education (PDIE) for its Hong Kong IPO on December 22, following approval from the Hong Kong Stock Exchange. China International Capital Corporation (CICC) serves as the sole sponsor.
As a leading Chinese AI company, Zhipu focuses on developing advanced general-purpose large models. Founded in 2019 with a vision to pursue artificial general intelligence (AGI) innovation, the company has consistently delivered cutting-edge AI research while expanding commercial applications to drive rapid revenue growth.
In 2021, Zhipu launched China’s first proprietary pre-trained large model framework, GLM, alongside its Model-as-a-Service (MaaS) platform for product development and commercialization. The following year, it open-sourced GLM-130B, a 100-billion-parameter model.
Zhipu operates in the large language model (LLM) segment, catering to institutional clients (private and public sector entities) and individual users (end-users and developers). As of June 30, 2025, its models supported over 8,000 institutional clients and approximately 80 million devices.
According to Frost & Sullivan, Zhipu ranked first among China’s independent general-purpose large model developers by 2024 revenue (6.6% market share) and second overall. Revenue surged from RMB 57.4 million in 2022 to RMB 312 million in 2024, achieving a CAGR exceeding 130%. However, losses widened due to heavy R&D investments, reaching RMB 2.96 billion in 2024 and RMB 2.36 billion in H1 2025.
**MaaS Platform and Deployment** Zhipu’s MaaS platform offers four model categories: language, multimodal, agent, and code models, plus tools for fine-tuning and deployment. Key advantages include comprehensive model coverage, scalability, and adaptability to diverse computing infrastructures.
Deployment options: - **On-premise**: Models hosted on client infrastructure, ideal for sensitive data and specialized use cases. - **Cloud**: Scalable solutions for agility, minimizing upfront infrastructure costs.
Revenue recognition varies by deployment: on-premise (upon delivery and acceptance) and cloud (subscription or usage-based).
**Core AI Capabilities** Zhipu’s models emulate human abilities: deep reasoning (reflective/ruminative models), world perception (multimodal models), and tool usage (agent models). Its flagship GLM-4.5 (355B parameters) ranked third globally and first in China per July 2025 benchmarks. The lightweight GLM-4.5-Air (106B parameters) followed.
Notable achievements: - **Global Rankings**: Fifth on ChatbotArena and WebDevArena (September 2025). - **Adoption**: Top 10 global token consumption on OpenRouter. - **Hallucination Rate**: Second-lowest globally in RAG benchmarks (September 2025).
Upgrades like GLM-4.6 (November 2025) further enhanced coding performance, topping CodeArena.
**Agent Models** AutoGLM, Zhipu’s base agent, bridges conversational AI to task execution, achieving state-of-the-art results on AgentBench. AutoGLM 2.0 (August 2025) enabled broader app/website automation. The advanced AutoGLM Rumination combines deep reasoning with operational capabilities for complex tasks.
**Market Outlook** China’s LLM market reached RMB 5.3 billion in 2024 (47B from institutions, 6B from individuals), projected to grow at a 63.5% CAGR to RMB 101.1 billion by 2030.
**Intellectual Property** Zhipu holds 86 Chinese patents (84 inventions), 160 copyrights, 314 trademarks, and 58 domains. Major shareholders include Meituan, Ant Group, Tsinghua Holdings, and Today Capital.
**IPO Proceeds Allocation** - 70%: R&D for general AI models. - 10%: MaaS platform optimization. - 10%: Partner network expansion and strategic investments. - 10%: Working capital.
Comments