Zhang Peng (Zhipu AI)
Last reviewed
Sources
26 citations
Review status
Source-backed
Revision
v3 · 2,382 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Sources
26 citations
Review status
Source-backed
Revision
v3 · 2,382 words
Add missing citations, update stale details, or suggest a clearer explanation.
Zhang Peng (Chinese: 张鹏) is a Chinese computer scientist and technology executive who serves as co-founder and chief executive officer of Zhipu AI, the Beijing-based developer of the GLM family of foundation models that has operated internationally as Z.ai since mid-2025. A longtime engineer from Tsinghua University's Knowledge Engineering Group (KEG), Zhang led the company from its registration as a university spinoff in June 2019 through its debut on the Hong Kong Stock Exchange on January 8, 2026, when it became, by its own description, the world's first publicly listed large language model company [1]. The listed entity trades as Knowledge Atlas Technology Joint Stock Co., Ltd. under stock code 2513, and by June 22, 2026 its market value had briefly topped HK$1 trillion (about USD 128 billion), a year-to-date gain exceeding 2,400 percent since the IPO [2][24].
Zhang belongs to the generation of Chinese AI founders who emerged from university laboratories rather than the consumer internet industry. He spent roughly two decades inside Tsinghua's KEG ecosystem, where he headed the work of turning the lab's knowledge graph and language model research into deployable systems [3]. At Zhipu he presides over a company that has positioned itself as one of China's most deliberate pursuers of artificial general intelligence (AGI), with a stated mission of making machines think like humans [1].
His tenure as CEO spans the defining events in the company's history: the open bilingual model GLM-130B in 2022, the ChatGLM era that followed ChatGPT, the company's addition to the United States Entity List in January 2025 [4], the Hong Kong initial public offering of January 2026 [1], and a follow-on application announced on June 1, 2026 to raise up to RMB 15 billion (about USD 2.2 billion) on the Shanghai Stock Exchange's STAR Market [5].
Zhang Peng is a common Chinese name. The Zhipu chief executive should not be confused with other public figures who share it, notably Zhang Peng, the founder and president of the Beijing technology media and events company GeekPark, who is also a prominent commentator on China's AI industry [6].
Zhang studied computer science as an undergraduate in Tsinghua University's Department of Computer Science and Technology, and later earned an engineering doctorate through the university's Innovation Leadership Project as part of its 2018 cohort [7]. Before founding Zhipu he served as deputy director of the Science and Technology Big Data Research Center at Tsinghua's Institute of Data Science [7].
His research career was rooted in KEG, the Tsinghua laboratory led by professors Tang Jie and Li Juanzi that specialized in knowledge graphs and data mining. As a principal researcher, Zhang contributed to the GLM series of large models as well as AMiner, the lab's academic knowledge graph and researcher-profiling system, and XLORE, a large bilingual knowledge graph, and published more than ten papers at venues including ICML and ISWC [7]. Within the lab he was the engineering counterpart to its academic leadership, responsible for converting research prototypes into working products [3].
Zhipu AI, formally Beijing Zhipu Huazhang Technology Co., Ltd., was registered in June 2019 as a spinoff of KEG. Zhang has recalled that the approval process took about 18 months beginning in 2017, and that the company was among the first to use a 2018 state policy that allowed in-service university researchers to commercialize their research results [3].
The founding team divided responsibilities along academic and operational lines. Tang Jie, who remained a Tsinghua professor, became the academic co-founder and chief scientist, and fellow KEG professor Li Juanzi is also credited as a co-founder. Liu Debing took the chairmanship with responsibility for strategy, while Zhang became CEO and general manager in charge of the business, research and development, and operations, with Wang Shaolan serving as president [8]. Tang Jie and Liu Debing have been identified as the company's key controlling shareholders [9].
Under Zhang's management Zhipu raised more than RMB 10 billion across successive rounds from corporate, financial, and state investors, including Alibaba, Tencent, Meituan, Xiaomi, Ant Group, HongShan, Qiming Venture Partners, and Legend Capital [9]. In mid-2024, Saudi Aramco's venture arm Prosperity7 Ventures joined a USD 400 million financing that valued the company at about USD 3 billion [10]. After the company was placed on the US Entity List, state capital became more prominent: in March 2025 Zhipu raised about USD 257 million in quick succession from government-backed funds in Hangzhou, Zhuhai, and Chengdu [11], and in July 2025 it secured roughly USD 140 million from Shanghai state funds, by which time its valuation had reached about RMB 40 billion (more than USD 5 billion) [12]. In July 2025 the company consolidated its international brand as Z.ai [13].
By September 2025 Zhipu reported more than 12,000 enterprise customers, with its models running on more than 80 million end-user devices [1]; Zhang has said its services are used by nine of China's ten largest internet companies [3]. The company's first annual results as a listed firm showed 2025 revenue of RMB 724 million, up 131.9 percent year on year at a gross margin of about 41 percent, against a net loss that widened to RMB 4.7 billion as model training and talent costs grew, with research and development spending of RMB 3.18 billion [5][25]. API revenue rose 293 percent and agent-business revenue 249 percent over the year [25].
Zhang has overseen the GLM (General Language Model) line since its origins in KEG research on autoregressive blank-infilling pretraining. Major releases under his leadership include:
| Model | Release | Notes |
|---|---|---|
| GLM-130B | 2022 | Open bilingual Chinese-English model with 130 billion parameters [14] |
| ChatGLM-6B | March 2023 | Open-weight conversational model released amid the post-ChatGPT boom; widely downloaded [14] |
| GLM-4 | January 2024 | Flagship model with tool-use and agent capabilities [14] |
| GLM-4.5 | July 2025 | 355-billion-parameter mixture-of-experts model under the MIT license, built for agentic tasks [15] |
| GLM-4.6 | September 2025 | Improved agentic coding model, roughly 15 percent more token-efficient than GLM-4.5 [16] |
| GLM-4.7 | December 2025 | Open-source release aimed at real development workflows [17] |
| GLM-5 | February 2026 | 744-billion-parameter open-weight MoE flagship under the MIT license [18] |
| GLM-5.1 | April 2026 | API launched March 27, 2026; open weights published April 7, 2026 [19] |
| GLM-5.2 | June 2026 | About 753-billion-parameter MoE (roughly 40 billion active per token) under the MIT license, tuned for long-horizon coding; reported SWE-bench Pro score of 62.1, ahead of GPT-5.5 at 58.6 [26] |
A distinctive feature of Zhang's strategy is releasing frontier-scale weights under the permissive MIT license while monetizing through APIs and an enterprise Model-as-a-Service (MaaS) platform. From GLM-4.5 onward, flagship weights have been distributed through Hugging Face and ModelScope alongside free consumer access on Z.ai [15][18]. The approach doubles as an ecosystem play and as a hedge against US export restrictions, and it has continued after the IPO: GLM-5 shipped as an open-weight model only a month after the listing, and GLM-5.2 followed in June 2026, with Z.ai releasing its core weights under an unrestricted MIT license on Hugging Face on the day of launch [18][26]. The company's planned STAR Market raise earmarks RMB 12 billion, 80 percent of the proceeds, for research and development of the next-generation GLM-6 and subsequent model series [5].
In mid-January 2025, the US Commerce Department's Bureau of Industry and Security added Beijing Zhipu Huazhang Technology and a group of affiliated entities to the Entity List, effective January 16, 2025, asserting that they advanced China's military modernization through the development and integration of advanced artificial intelligence research [20]. The action, taken in the final days of the Biden administration, made Zhipu the first of China's prominent large model startups to be blacklisted [4]. The company responded that the decision "lacks factual basis," said it firmly opposed the designation, and stated that inclusion would not materially affect its business, adding that it did not rely on American large-model technology and would continue to participate in global AI competition [21].
The sanction did not stop the listing drive. In April 2025 Zhipu registered for onshore IPO counseling with China International Capital Corporation as sponsor, the first of China's "AI tiger" model startups, a group that also includes Moonshot AI, MiniMax, Baichuan, StepFun, and 01.AI, to formally start a listing process [9]. The company ultimately adopted a Hong-Kong-first path. On January 8, 2026 it listed on the Hong Kong Stock Exchange at an issue price of HK$116.20 per share, opening at HK$120.00 for an opening market capitalization of HK$52.83 billion and raising about HK$4.2 billion; the stock closed its first day up about 13 percent [1][5][22]. The retail tranche was reported to be oversubscribed by more than 1,000 times [22]. The debut came one day before MiniMax's, making Zhipu the first of the AI tigers to go public [2][22]. At the ceremony Zhang called the listing "a landmark milestone" and said the company would keep working to make machines think like humans while shouldering greater social responsibility [1].
Zhang has framed the IPO as a means rather than an end, comparing the company's path to a marathon in which founders "have to persist" and treat capital events as supply stations where one can "replenish yourself" mid-race [3].
On June 1, 2026, less than five months after the Hong Kong debut, Zhipu announced that it would apply for an A-share listing on Shanghai's STAR Market, seeking up to RMB 15 billion, of which RMB 12 billion is allocated to foundation model research including GLM-6, RMB 2 billion to its MaaS platform, and RMB 1 billion to working capital. MiniMax filed its own A-share tutoring papers on May 29, 2026, positioning the two firms to become the first pure-play large language model companies on the A-share market [5]. On June 17, 2026 the company's STAR Market IPO tutoring status was updated to "tutoring acceptance," with Guotai Haitong Securities as tutoring institution [24]. As Hong Kong investors bid the stock higher through the spring, the Hong Kong shares topped a HK$1 trillion market capitalization on June 22, 2026, briefly reaching about HK$2,980 intraday before easing [24].
Zhang describes Zhipu's goal as "super cognitive intelligence beyond human level" and has argued that AGI "is not just a ChatBot, a language model, or just the number of model parameters" [23]. He has laid out a staged view of AI capability that progresses from language ability through logical reasoning, tool use, and self-learning [23]. At the WISE2024 conference in November 2024, he said it was "too early" to worry that the scaling law had hit a ceiling, pointing to unexplored headroom in vision, hearing, and other modalities [23].
On the China-US rivalry and AI investment, Zhang has dismissed bubble concerns, arguing that China's AI investment remains far below what the technology requires, "only a fraction, maybe one-tenth," and asking, "If we don't invest, will it just naturally happen?" [3]. After the Entity List designation he maintained that Zhipu's technology stack did not depend on the United States [21]. Amid China's model price wars following the IPO, he defended the company's pricing by arguing that "price is determined by value," emphasizing the value delivered per token to customers [5].