Z.ai
Z.ai builds the GLM large language models and sells model access through a developer API platform.
Founded
2019Z.ai builds the GLM large language models and sells model access through a developer API platform.
Z.ai raised roughly US$5 billion in September 2026 through two concurrent instruments: a placement of about 21.97 million new H shares at HK$714 and zero-coupon convertible bonds of RMB 20.14 billion convertible at HK$892.50 and maturing in September 2027. Sources reported six institutional placees without naming them, and no source named the bondholders.
The same filings and reports describe use-of-proceeds weighting toward next-generation models and self-training systems. The September transactions follow a January 2026 initial public offering and a July 2026 placement, so the entity now shows three completed capital raises since listing.
Z.ai's core offering is the GLM (General Language Model) series of large language models, distributed to developers through its model platform. GLM-5 was released in February 2026 and GLM-5.3 followed in August 2026, presented by the company as its frontier coding release, with a lighter GLM-5.3-Flash variant offered at lower token prices.
Beyond the base models, the company ships the Z.ai consumer chat and agent application, the ZCode coding application released in July 2026, the AutoGLM agent framework, CodeGeeX, CogView, and the developer documentation and pricing surface at docs.z.ai.
Z.ai became the first standalone large-model company to list on a public exchange when it completed its Hong Kong offering in January 2026, and it followed that debut with two large equity taps, an approximately HK$31.41 billion placement in July 2026 and a combined share and convertible bond raise of about US$5 billion announced in September 2026. Its market capitalisation crossed HK$1 trillion around a June 2026 model launch.
Management has pointed to a secondary Shanghai listing as a further capital source. Investor attention now centres on whether platform and coding revenue can absorb the research spending that the company continues to prioritise, with most disclosed proceeds earmarked for next-generation model work.
Zhipu AI benefits from deep academic roots at Tsinghua University, where its founding team including Chief Scientist Tang Jie and CEO Zhang Peng conducted foundational research. The company maintains strong government relationships and positions itself as a domestic full-stack AI provider with intellectual property trained on Huawei Ascend chips.
The enterprise-first strategy has attracted over 12,000 clients, while the developer ecosystem exceeds 45 million users. Zhipu AI became the first foundation model company to list on a public exchange in China, providing access to public capital markets for continued research investment.
Z.ai's losses remain large relative to its revenue. Reported first-half 2026 revenue rose about fourfold to roughly RMB 953.9 million while the company still recorded a net loss of about RMB 2 billion, so model training and infrastructure spending continue to outpace commercialisation.
The company also carries constraints its American rivals do not. It was added to the US Entity List in January 2025, limiting access to certain US technology, and its share price fell sharply from its 2026 highs before its September 2026 fundraising, during which management said capital would keep flowing to research and self-training systems.
Z.ai sells model access through per-token API pricing on its developer platform and through subscription tiers on its ZCode coding product, with the GLM Coding Plan offered in Lite, Pro and Max levels. As of the GLM-5.3 release window the company priced GLM-5.1 at $1.40 per million input tokens and $4.40 per million output tokens, with the Flash variant discounted sharply below that.
This two-track structure lets the company keep flagship models at premium rates while using the Flash tier to compete on cost, a posture consistent with the price competition among Chinese foundation-model vendors through 2024, 2025 and 2026.