Company intelligence

Z.ai

Z.ai (formerly Zhipu AI) is a Beijing-based AI lab developing the GLM family of large language models and agentic developer tools like ZCode.

AI infraPublic · 2513.HKBeijing, ChinaFounded 2019Founders: Tang Jie, Li JuanziWebsite ↗See connections →
Last verified Jan 2026

Z.ai at a glance

Founded
2019[1]
Headquarters
Beijing, China[2]
  1. [1] Founded: Wikipedia (retrieved September 5, 2026)
  2. [2] Headquarters: Wikipedia (retrieved September 5, 2026)

Z.ai profile

written 2026-11-07

Z.ai, formerly Zhipu AI, is a Beijing lab behind the GLM models that listed in Hong Kong in January 2026 at HK$116.20 a share, about a $6.7 billion valuation.

What Z.ai does

Z.ai is a Chinese artificial intelligence developer that builds the GLM family of large language models. TechCrunch reported in 2023, when the company still went by its original name Zhipu AI, that it was founded in 2019 and was spun out of Tsinghua University, led by Tang Jie, a professor in the university's Department of Computer Science and Technology.[1]

Z.ai's current flagship model is GLM-5.3, an open-source large language model the company released in mid-August 2026, according to SiliconANGLE. The underlying architecture, shared with the earlier GLM-5.2 release, is a mixture-of-experts design with 753 billion parameters and a context window of 1 million tokens; GLM-5.3 kept that design but went through a more extensive post-training process.[2]

Z.ai says GLM-5.3 is especially capable at finding software vulnerabilities. SiliconANGLE reported that the model has so far found more than 2,400 vulnerabilities in 269 software projects, about half of them rated medium severity or higher, and that one flaw it found was in code written 40 years ago. On CyberGym, a benchmark that tests an LLM's ability to find code vulnerabilities, GLM-5.3 outperformed Anthropic's Claude Mythos 5, though it trailed that model on two other cybersecurity benchmarks.[2]

Z.ai trained GLM-5.3 using sandboxes meant to mimic real developer workstations. SiliconANGLE reported that the company generated these sandboxes with specialized AI agents that modeled the environments on real-world software projects, created coding exercises for each one, and used a separate 'judge agent' to confirm each exercise was solvable before GLM-5.3 attempted it; some exercises took days to complete, which the company says improved the model's ability to handle long-horizon tasks. GLM-5.3 is available through Z.ai's GLM Coding Plan subscription, and the company said it planned to publish the model's weights on Hugging Face under an open-source license within two weeks of the release.[2]

Funding

Z.ai, then named Zhipu AI, told TechCrunch in October 2023 that it had raised 2.5 billion yuan ($340 million) in total financing to date that year. The money came from yuan-denominated funds rather than the dollar funds that had traditionally financed Chinese tech, a shift TechCrunch tied to new U.S. restrictions on American investment in China's AI, semiconductor and quantum-computing sectors.[1]

Zhipu's 2023 investor roster, as reported by TechCrunch, included HongShan (the renamed Sequoia Capital China), Shunwei Capital, Hillhouse Capital and a state fund managed by Legend Capital, alongside Chinese internet companies Alibaba, Tencent, Ant Group, Xiaomi, Meituan, Kingsoft, TAL Education Group and Boss Zhipin.[1]

Before listing, Zhipu had raised roughly US$1.5 billion since founding from investors including Alibaba, Tencent, Ant Group, Meituan, Xiaomi, Hillhouse, Qiming Venture Partners, Chinese local-government funds and Saudi Aramco's Prosperity7 Ventures, The AI Rankings reported, citing Bloomberg.[3]

Z.ai listed on the Hong Kong Stock Exchange on January 8, 2026, under stock code 2513, billing itself as the world's first listing by a large-language-model company. Its shares were priced at HK$116.20, and the offering raised about $558 million to $560 million, implying a valuation of approximately $6.7 billion, according to China Internet Watch and Crypto Briefing. Retail investors oversubscribed the offering 1,159 times, and the stock closed 13.2% higher at HK$131.50 on its debut.[3],[4],[5]

Z.ai has said it is directing proceeds toward research and development, with an emphasis on computing capacity and its next-generation GLM infrastructure, according to Crypto Briefing.[5]

History

Z.ai started life as Zhipu AI, founded in 2019 as a spinout of Tsinghua University, according to TechCrunch's 2023 reporting. The company's early public work included an open-sourced bilingual Chinese-English conversational model, ChatGLM-6B, trained on six billion parameters, plus a larger open-sourced foundation model, GLM-130B, trained on 130 billion parameters, both described in the same TechCrunch article.[1]

Z.ai rebranded from the Zhipu AI name at some point before mid-2026; Fortune's July 2026 reporting and SiliconANGLE's August 2026 reporting both refer to the company only as Z.ai. The sources opened for this profile do not state the exact date of the rebrand, so none is given here.[6],[2]

Z.ai released GLM-5.2, a 753-billion-parameter mixture-of-experts model, in mid-July 2026, according to SiliconANGLE. GLM-5.3 followed on August 14, 2026, using the same architecture after what the company described as more extensive post-training.[2]

Competitors

Z.ai competes with U.S. foundation-model labs on coding and cybersecurity benchmarks. SiliconANGLE reported that GLM-5.3 outperformed Anthropic's Claude Mythos 5 on CyberGym, a vulnerability-discovery benchmark, while falling behind the same Anthropic model on two other cybersecurity benchmarks tested.[2]

Z.ai is one of several Chinese AI labs that went to Hong Kong's stock market in the same period as it built out GLM. Fortune's July 2026 coverage of 'the Great Chinese AI IPO Rush' grouped Z.ai with Minimax, which debuted alongside it in January, and noted that Moonshot AI, maker of the Kimi assistant, was separately preparing its own Hong Kong listing at a valuation of just over $30 billion.[6]

Business and financials

Z.ai's revenue is enterprise-heavy. China Internet Watch reported, citing the prospectus, that Zhipu generated 85% of its first-half 2025 revenue from selling on-premise model deployments to state-owned enterprises and financial institutions, and reported overall gross margins of 50% for that half.[4],[2]

Its first annual results as a listed company reported 2025 revenue of about RMB 724 million, up roughly 132% year on year, against a net loss of about RMB 4.7 billion, according to The AI Rankings; AITool reported the same figures.[3],[7]

In January 2025 Zhipu was added to the US Entity List, which restricts American companies from supplying certain technology and components to the firm, Crypto Briefing reported.[5]

Frequently asked questions

Is Z.ai the same company as Zhipu AI? Yes. Zhipu AI, the Tsinghua University spinout founded in 2019, now operates under the Z.ai brand and trades in Hong Kong as Knowledge Atlas Technology, stock code 2513.[5],[3],[1]

When did Z.ai go public? On January 8, 2026, on the Hong Kong Stock Exchange, at HK$116.20 a share.[3],[4]

Who are Z.ai's investors? Pre-IPO backers include Alibaba, Tencent, Ant Group, Meituan, Xiaomi, Hillhouse, Qiming Venture Partners and Prosperity7 Ventures.[3]

Is something wrong on this page?

Company Snapshot

Z.ai is an AI infra company, based in Beijing, China, founded 2019 by Tang Jie, Li Juanzi. Total funding ~$1.5B (pre-IPO, since founding) (As of Jan 2026).

Recent changes

  1. Jul 2, 2026Z.ai Launches ZCode, a Free IDE for GLM-5.2, to Challenge Cursor, Claude Code and Copilot Pulse story
  2. Jul 2, 2026Chinese Lab Z.ai Launches ZCode to Challenge Cursor, Claude Code and GitHub Copilot Pulse story
  3. Jan 8, 2026IPO ~$558M at ~$6.7B Crypto Briefing
  4. Oct 20, 2023First reported total funding: $340M TechCrunch

Debt, IPO & other events

not counted in equity raised
DateEventAmountValuationInvestorsSource
Jan 8, 2026IPO~$558M~$6.7B
Crypto Briefing

Investors

Latest Pulse

  1. GLM-5.3 Hits the API at $1.40/$4.40 Per Million TokensZ.ai priced its new GLM-5.3 model identically to GLM-5.2 at $1.40 per million input tokens and $4.40 per million output tokens, undercutting Grok, Kimi, Claude Opus and GPT on a simple blended-cost comparison.→
  2. Z.ai's GLM-5.3 Takes Aim at Claude and GPT in CodingChinese AI lab Z.ai is upgrading its flagship GLM-5.3 model with sharper coding capabilities, the latest sign that Chinese labs are closing the gap with Anthropic and OpenAI on developer-focused benchmarks.→
  3. China's Z.ai Holds Back Its Best Model Over Hacking RiskZ.ai built GLM-5.3 into the strongest open coding and vulnerability-hunting model it has released, then delayed publishing its weights because its own cyber capability tests came back higher than the company expected.→
  4. China's AI Labs Race Toward A Wave Of IPOsMoonshot, DeepSeek, MiniMax and Z.ai are all pursuing or completing public listings within months of each other, a compressed wave of Chinese AI IPOs that resets how investors value the entire sector.→
  5. 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.→
  6. Chinese Lab Z.ai Launches ZCode to Challenge Cursor, Claude Code and GitHub CopilotZ.ai launched ZCode on July 2, an agentic development environment built around its GLM Coding Plan and long-horizon task execution — the user describes an outcome, the agent plans the work, edits files, runs checks and iterates until the goal is met. It enters a coding-agent market where, per VentureBeat, virtually every developer shipping software in 2026 is already using Claude Code, GitHub Copilot or Cursor, positioning ZCode as a lower-cost, desktop-native challenger.→

Coverage Timeline

newest first
  1. Aug 19, 2026AIGLM-5.3 Hits the API at $1.40/$4.40 Per Million Tokens→
  2. Aug 16, 2026AIZ.ai's GLM-5.3 Takes Aim at Claude and GPT in Coding→
  3. Aug 14, 2026AIChina's Z.ai Holds Back Its Best Model Over Hacking Risk→
  4. Jul 23, 2026IPOChina's AI Labs Race Toward A Wave Of IPOs→
  5. Jul 2, 2026AIZ.ai Launches ZCode, a Free IDE for GLM-5.2, to Challenge Cursor, Claude Code and CopilotUp to 82% cheaper than Opus 4.8→
  6. Jul 2, 2026AIChinese Lab Z.ai Launches ZCode to Challenge Cursor, Claude Code and GitHub CopilotNew product launch→

All Coverage

AI
GLM-5.3 Hits the API at $1.40/$4.40 Per Million Tokens
Illustration for: GLM-5.3 Hits the API at $1.40/$4.40 Per Million Tokens
AI

GLM-5.3 Hits the API at $1.40/$4.40 Per Million Tokens

Z.ai priced its new GLM-5.3 model identically to GLM-5.2 at $1.40 per million input tokens and $4.40 per million output tokens, undercutting Grok, Kimi, Claude Opus and GPT on a simple blended-cost comparison.

DEEP DIVE
AI
Z.ai's GLM-5.3 Takes Aim at Claude and GPT in Coding
AI

Z.ai's GLM-5.3 Takes Aim at Claude and GPT in Coding

Chinese AI lab Z.ai is upgrading its flagship GLM-5.3 model with sharper coding capabilities, the latest sign that Chinese labs are closing the gap with Anthropic and OpenAI on developer-focused benchmarks.

DEEP DIVE
AI
China's Z.ai Holds Back Its Best Model Over Hacking Risk
AI

China's Z.ai Holds Back Its Best Model Over Hacking Risk

Z.ai built GLM-5.3 into the strongest open coding and vulnerability-hunting model it has released, then delayed publishing its weights because its own cyber capability tests came back higher than the company expected.

DEEP DIVE
IPO
China's AI Labs Race Toward A Wave Of IPOs
IPO

China's AI Labs Race Toward A Wave Of IPOs

Moonshot, DeepSeek, MiniMax and Z.ai are all pursuing or completing public listings within months of each other, a compressed wave of Chinese AI IPOs that resets how investors value the entire sector.

AI
Z.ai Launches ZCode, a Free IDE for GLM-5.2, to Challenge Cursor, Claude Code and Copilot
Illustration for: Z.ai Launches ZCode, a Free IDE for GLM-5.2, to Challenge Cursor, Claude Code and Copilot
AIUp to 82% cheaper than Opus 4.8

Z.ai Launches ZCode, a Free IDE for GLM-5.2, to Challenge Cursor, Claude Code and Copilot

Z.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.

AI
Chinese Lab Z.ai Launches ZCode to Challenge Cursor, Claude Code and GitHub Copilot
Illustration for: Chinese Lab Z.ai Launches ZCode to Challenge Cursor, Claude Code and GitHub Copilot
AINew product launch

Chinese Lab Z.ai Launches ZCode to Challenge Cursor, Claude Code and GitHub Copilot

Z.ai launched ZCode on July 2, an agentic development environment built around its GLM Coding Plan and long-horizon task execution — the user describes an outcome, the agent plans the work, edits files, runs checks and iterates until the goal is met. It enters a coding-agent market where, per VentureBeat, virtually every developer shipping software in 2026 is already using Claude Code, GitHub Copilot or Cursor, positioning ZCode as a lower-cost, desktop-native challenger.

Related Companies

More on Z.aiAI Buildout Tracker →Big Tech AI Spending →AI Valuations →

Z.ai FAQ

When was Z.ai founded and where is it based?

Z.ai was founded in 2019 and is headquartered in Beijing, China, according to Wikipedia.

Sources & methodologyshow
Where these figures come from

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 ·