Company intelligence

Zhipu AI

Chinese foundation model developer building the GLM family of large language models, offering them via API and cloud/on-prem enterprise deployments.

AI infraPublic · 2513.HKBeijing, ChinaFounded 2019Public companyWebsite ↗
Last verified Jan 8, 2026

Zhipu AI at a glance

Founded
2019[1]
Headquarters
Beijing, China[2]
  1. [1] Founded: Zhipu AI (via Bing search result) (retrieved September 5, 2026)
  2. [2] Headquarters: CNBC (retrieved January 8, 2026)

Zhipu AI profile

written 2026-11-19

Zhipu (Z.ai) is a Beijing-based large language model developer that listed in Hong Kong on Jan. 8, 2026 under ticker 2513.

Who they are

Zhipu is a Beijing-based developer of large language models. SiliconANGLE describes the company as formerly known as Zhipu AI and officially known as Knowledge Atlas Technology Joint Stock Co. Ltd. It goes by Z.ai internationally, and its website carries that name.[1],[2]

The company was founded in 2019 by researchers from Tsinghua University, according to Jing Daily. It reported revenue of 312.4 million RMB in 2024, a figure Jing Daily gives as about $44.7 million.[3]

CNBC reports that the firm has offices in the United Kingdom, Singapore, Malaysia and across the Middle East. It also runs joint innovation centers in Southeast Asia, including in Indonesia and Vietnam. CNBC describes the company as strongly backed by Beijing and as one of China's so-called AI tigers, a group of startups building language models to rival those of OpenAI and Anthropic.[4]

Products and customers

Zhipu is best known for its GLM series of pretrained language models, which handle conversation in Chinese and English. SiliconANGLE also lists a video generation model called Ying and an agent application called AutoGLM, which can run on smartphones and use third-party apps and tools to automate tasks for users.[1]

Z.ai's developer documentation says it offers a variety of models and agents for different scenarios. The documentation highlights open-source coding capabilities among its core models.[5]

In its IPO filing, the company said it expects most of its sales to come from enterprise customers. It follows a model-as-a-service approach, and SiliconANGLE reports that its API platform serves more than 2.7 million enterprise and application developers. SiliconANGLE also notes that the company has received significant funding from Chinese technology groups such as Tencent and Alibaba in the past.[1]

Hong Kong listing

Zhipu listed on the Hong Kong stock exchange on Jan. 8, 2026 under the ticker 2513. SiliconANGLE reported ahead of the listing that the company described itself as the world's first publicly listed foundation model company. CNBC reports that the offering was a $558 million initial public offering and that it made Zhipu the first of China's AI tigers to go public.[1],[4]

According to SiliconANGLE, the company planned to issue 37,419,500 class H shares at an offer price of HK$116.20. CNBC reports that the stock closed its first day up 13.1%.[1],[4]

The prospectus allocates the proceeds by use. SiliconANGLE reports that 70% goes to research and development, with 10% each for commercial offerings and the partner ecosystem, and the remaining 10% for working capital and other corporate expenses.[1]

Financials and constraints

SiliconANGLE reports figures from the IPO filing. In the first half of 2024 the company had revenue of $6.4 million and an operating loss of $146.9 million. In the first half of 2025 revenue rose to $27.2 million and losses grew to $271.1 million. Research and development costs were $122.5 million in the first half of 2024 and $227.5 million in the first half of 2025.[1]

Zhipu was added to the US Commerce Department's Entity List in January 2025, after US officials said it was working with China's military, according to CNBC. CNBC adds that the firm's ability to train its models has also been constrained by US restrictions on access to advanced semiconductor technology and expertise.[4]

Is something wrong on this page?

Company Snapshot

Zhipu AI is an AI infra company, based in Beijing, China, founded 2019. Total funding $558M (Hong Kong IPO) (As of Jan 8, 2026).

Recent changes

  1. Jan 8, 2026IPO $558M CNBC
  2. Jan 8, 2026First reported total funding: $558M (Hong Kong IPO) CNBC

Debt, IPO & other events

not counted in equity raised
DateEventAmountValuationInvestorsSource
Jan 8, 2026IPO$558M—
CNBC

Latest Pulse

  1. Cohere Targets CAD $35.7B Valuation In New Series ECohere announced the initial close of a Series E round targeting up to CAD $7 billion that would value it at CAD $35.7 billion, led by Germany's Schwarz Group.→
  2. DeepSeek Raises $15B Days After US RestrictionsDeepSeek has raised about $15 billion (100 billion yuan) in a new round, Bloomberg reported, two days after the US ordered federal agencies to remove its technology.→
  3. Tsinghua Professor's Stealth LLM Startup Hits $1.4BA stealth-mode Chinese LLM startup founded in February by a Tsinghua University professor has quietly reached a $1.4 billion valuation after raising $400 million across three rounds, including from Tencent.→
  4. Moonshot AI Confidentially Files for $3B HK IPOMoonshot AI, maker of the Kimi chatbot, confidentially filed to list on the Hong Kong Stock Exchange and is aiming to raise $3 billion, Reuters reported; the company has separately been valued at $50 billion in an ongoing private funding round, even as US officials weigh sanctions over its Nvidia chip use.→
  5. China Weighs Export Controls on AI Models, ChipsChina's Ministry of Commerce is weighing export controls on advanced AI models, training data and chipmaking access, including barring TSMC from producing chips based on Chinese company designs.→
  6. Goldman Names Its Preferred Chinese AI ModelsGoldman Sachs research named Zhipu, DeepSeek and ByteDance its preferred Chinese AI models, ranked ahead of Alibaba and Tencent on speed-to-market, and recommended rotating out of Korean AI stocks into a long China AI basket.→
  7. Tencent's Hy3 Takes on GLM-5.2 at Half the SizeTencent's Apache-licensed Hy3 model wins most benchmarks against GLM-5.2 at roughly half the parameter count, though it still trails on coding tasks, intensifying China's open-weight model race.→
  8. Tencent's Open Hy3 Model Beats GLM-5.2 at Half the SizeTencent released Hy3, an Apache-licensed open-weight model roughly half the parameter count of Zhipu's GLM-5.2, claiming wins across most benchmark categories except coding tasks -- intensifying the open-model race among Chinese AI labs.→
  9. Z.ai Launches ZCode, a Free IDE for GLM-5.2, to Challenge Cursor, Claude Code and CopilotZ.ai, the Beijing lab formerly known as Zhipu AI, officially launched ZCode on July 2 — a free, agent-first development environment purpose-built for its GLM-5.2 model, with subscription plans starting at $16.20/month, up to 82% cheaper than Anthropic's Claude Opus 4.8 API pricing. The launch lands three weeks after a US export-control order briefly suspended foreign access to Anthropic's Claude Fable 5 and Mythos 5, an episode Z.ai's timing has directly capitalized on, with GLM-5.2 running entirely on Huawei silicon and ranking second globally on the Code Arena leaderboard behind only Claude Fable 5.→
  10. GLM-5.2 Becomes the Leading Open-Weights Model on Artificial AnalysisGLM-5.2 took the top spot among open-weights models on the Artificial Analysis Intelligence Index, narrowing the gap to the closed frontier labs. For builders, a state-of-the-art model you can self-host is the strongest hedge yet against the regulatory and vendor risk that just took down Fable 5.→

Coverage Timeline

newest first
  1. Jul 12, 2026AIGoldman Names Its Preferred Chinese AI ModelsPT HK$1,880 (Zhipu)→
  2. Jul 6, 2026AITencent's Hy3 Takes on GLM-5.2 at Half the SizeWins most benchmarks at ~50% size→
  3. Jul 6, 2026AITencent's Open Hy3 Model Beats GLM-5.2 at Half the SizeHalf-size open model, Apache 2.0→
  4. Jun 16, 2026AIGLM-5.2 Becomes the Leading Open-Weights Model on Artificial Analysis→

All Coverage

AI
Goldman Names Its Preferred Chinese AI Models
Illustration for: Goldman Names Its Preferred Chinese AI Models
AIPT HK$1,880 (Zhipu)

Goldman Names Its Preferred Chinese AI Models

Goldman Sachs research named Zhipu, DeepSeek and ByteDance its preferred Chinese AI models, ranked ahead of Alibaba and Tencent on speed-to-market, and recommended rotating out of Korean AI stocks into a long China AI basket.

AI
Tencent's Hy3 Takes on GLM-5.2 at Half the Size
Illustration for: Tencent's Hy3 Takes on GLM-5.2 at Half the Size
AIWins most benchmarks at ~50% size

Tencent's Hy3 Takes on GLM-5.2 at Half the Size

Tencent's Apache-licensed Hy3 model wins most benchmarks against GLM-5.2 at roughly half the parameter count, though it still trails on coding tasks, intensifying China's open-weight model race.

AI
Tencent's Open Hy3 Model Beats GLM-5.2 at Half the Size
Illustration for: Tencent's Open Hy3 Model Beats GLM-5.2 at Half the Size
AIHalf-size open model, Apache 2.0

Tencent's Open Hy3 Model Beats GLM-5.2 at Half the Size

Tencent released Hy3, an Apache-licensed open-weight model roughly half the parameter count of Zhipu's GLM-5.2, claiming wins across most benchmark categories except coding tasks -- intensifying the open-model race among Chinese AI labs.

AI
GLM-5.2 Becomes the Leading Open-Weights Model on Artificial Analysis
Illustration for: GLM-5.2 Becomes the Leading Open-Weights Model on Artificial Analysis
AI

GLM-5.2 Becomes the Leading Open-Weights Model on Artificial Analysis

GLM-5.2 took the top spot among open-weights models on the Artificial Analysis Intelligence Index, narrowing the gap to the closed frontier labs. For builders, a state-of-the-art model you can self-host is the strongest hedge yet against the regulatory and vendor risk that just took down Fable 5.

Value Add Research

Related Companies

More on Zhipu AIAI Buildout Tracker →Big Tech AI Spending →AI Valuations →

Zhipu AI FAQ

When was Zhipu AI founded and where is it based?

Zhipu AI was founded in 2019 and is headquartered in Beijing, China, according to Zhipu AI (via Bing search result) and CNBC.

Sources & methodologyshow
Where these figures come from
  • CNBCmajor outlet2 facts

Figures are labelled Reported (a source states it), Estimated, Derived (computed from other facts) or Rumored (in talks / unconfirmed). Rumors never replace a confirmed figure. Each figure keeps its own as-of date; history is preserved rather than overwritten.

How we verify data ·