Z.ai

Corporate brand

Unclaimed Read 4× by AI crawlers Request correction

Chinese artificial-intelligence company developing large language models, multimodal systems, AI agents, and related services.

Last updated October 14, 2026

AI crawler reads

4requests

1companies

Amazon4

Profile: 2
Controversies: 2
JSON · MD: 0

First read: Oct 29, 2026
Latest read: Nov 3, 2026

Since Oct 28, 2026 · from this site's server log

Overview

Z.ai, formerly known internationally as Zhipu AI, is a Beijing-based Chinese artificial-intelligence company focused on large language models and related generative-AI products. Its Chinese corporate name is Beijing Zhipu Huazhang Technology Co., Ltd., while its international brand changed to Z.ai in 2025. The company developed from research activity associated with Tsinghua University and became an independent startup in 2019. Its work is closely connected with the General Language Model, or GLM, research program, which was developed by researchers at Tsinghua's Knowledge Engineering Group and Zhipu AI. The GLM family is the company's central technology platform. Early GLM research used an autoregressive blank-infilling approach, in which sections of text are removed and the model learns to reconstruct them while also supporting conventional language generation. The resulting model family has included conversational systems, general-purpose language models, vision-language models, speech systems, and text-to-video technology. Products and services have been aimed at individual users, developers, enterprises, public-sector customers, and research communities. Z.ai expanded rapidly during China's generative-AI investment cycle. In 2023, it reportedly received financing from a group that included Alibaba Group, Tencent, Meituan, Ant Group, Xiaomi, and HongShan. Prosperity7 Ventures participated in a further financing round announced in 2024. The company has been described by industry analysts and investors as one of China's leading so-called AI tiger companies and as a significant participant in the Chinese large-language-model market. Its product portfolio includes the GLM model series, the Ying text-to-video system, AutoGLM smartphone-agent software, and AMiner, a scholarly and research-information database created by Jie Tang and later owned by Z.ai. The company has also pursued international distribution and infrastructure compatibility, including work involving Huawei Ascend processors and overseas offices or innovation projects. Z.ai's international position has been shaped by both technological competition and geopolitical restrictions. In January 2025, the United States Commerce Department added the company to the Entity List, citing national-security concerns. In 2025, Z.ai began releasing selected GLM models under the MIT License, emphasizing open weights and wider developer access. It also continued model upgrades, including GLM-4.5, GLM-4.6, GLM-4.7, and the GLM-5 generation. The company completed an initial public offering on the Hong Kong Stock Exchange on 8 January 2026, becoming one of the first major Chinese large-language-model companies to reach the public markets. Later in 2026, it released GLM-5 and successive updates, while facing reported compute shortages, user-access restrictions, pricing changes, and competition from leading United States model providers. Z.ai therefore combines a research-oriented origin with a commercial strategy spanning foundation models, APIs, applications, open-weight releases, enterprise services, and international expansion. Z.ai is a Chinese artificial-intelligence company whose principal business is the research, development, and commercialization of large language models. The company was founded in 2019 by researchers connected with Tsinghua University and initially operated under the Zhipu AI name. Its international brand changed to Z.ai in 2025. The company’s domestic legal identity has been reported as Beijing Zhipu Huazhang Technology Co., Ltd., while its listed corporate name is Z.AI Co., Ltd. The company’s central technology family is GLM, short for General Language Model. The research behind GLM was developed with Tsinghua’s Knowledge Engineering Group and was described in a 2022 Association for Computational Linguistics paper. The approach uses an autoregressive blank-infilling method: portions of text are removed and the model learns to reconstruct them while also handling conventional language-generation tasks. The resulting product line, initially associated with ChatGLM, expanded into increasingly capable general-purpose language models, voice systems, vision-language models, and software agents. Z.ai has positioned itself as one of China’s leading domestic alternatives to US-developed foundation-model providers. Investors and industry observers have grouped it among China’s major AI “tiger” companies, and the International Data Corporation has described it as one of the country’s largest large-language-model participants. Its commercial activities include model access through APIs and subscriptions, enterprise AI deployments, open-weight model releases, research tools, and government or city-level projects. The company also owns AMiner, a scholarly and research-information database founded by Jie Tang in 2006. Capital raised from major Chinese technology and internet companies helped fund the company’s expansion. Reported investors have included Alibaba Group, Tencent, Meituan, Ant Group, Xiaomi, and HongShan. In 2024, Prosperity7 Ventures participated in a financing round reported at approximately US$400 million, with a valuation of about US$3 billion. In 2025, the company disclosed preparations for a public offering and completed a reported 61.28 million yuan government contract connected with city projects in Hangzhou. Z.ai’s model strategy has increasingly emphasized open weights. In July 2025, the GLM-4.5 and GLM-4.5 Air models were released under the MIT License, followed by subsequent GLM releases and vision-language systems. The company has worked to make its models compatible with Chinese computing infrastructure, including Huawei Ascend processors and chips from Cambricon Technologies and Moore Threads. Its broader product portfolio includes Ying, a text-to-video model, AutoGLM, a voice-operated mobile agent, and GLM-4-Voice, an end-to-end speech model capable of adjusting tone and dialect. The company entered the public markets in January 2026 through an initial public offering on the Hong Kong Stock Exchange, described in the reference material as the first IPO by a major Chinese LLM company. Later in 2026 it released GLM-5, GLM-5.1, GLM-5.2, and GLM-5.3, with reported improvements in context length, post-training, coding capability, and support for domestic chips. At the same time, Z.ai faced operational pressure when a shortage of computing resources caused service complaints, restricted new registrations, and contributed to a reported fall in its share price. It subsequently raised API prices while maintaining a lower reference price than Anthropic’s cited Opus 4.6 model. The company therefore combines an open-model research identity with the infrastructure, pricing, and capacity challenges typical of a rapidly scaling AI platform.

History

Z.ai originated in research associated with Tsinghua University and became an independent company in 2019. Its technical foundation is the GLM family, developed through collaboration involving Zhipu AI and Tsinghua's Knowledge Engineering Group. A 2022 academic publication described GLM's autoregressive blank-infilling method, which combines text reconstruction with language generation. The company entered a period of rapid expansion in 2023. It raised substantial financing from major Chinese technology and consumer-platform investors, while its models became part of the competition to establish domestic alternatives to leading United States AI systems. In 2024, Zhipu announced work on Sora-like video-generation technology, released the Ying text-to-video model, launched GLM-4-Plus, and introduced GLM-4-Voice, an end-to-end speech model capable of varying tone and dialect. In 2025, the company prepared for a public offering and began releasing selected models under the MIT License. It also won a Chinese government-related city-project contract in Hangzhou. During the same year, it released GLM-4.5, GLM-4.5 Air, GLM-4.5V, GLM-4.6, and later GLM-4.6V and GLM-4.7. The company adopted Z.ai as its international brand and worked to make its models compatible with domestic Chinese processors, including Huawei Ascend and chips from Cambricon Technologies. Its international activities expanded to offices and innovation projects in several overseas markets. The company faced geopolitical pressure when the United States Commerce Department placed it on the Entity List in January 2025. Despite this restriction, Z.ai continued model development and international positioning around open-weight systems. It also owned AMiner, a scholarly research database founded by Jie Tang in 2006. On 8 January 2026, Z.ai completed an initial public offering on the Hong Kong Stock Exchange. It subsequently released GLM-5 and later GLM-5.1, GLM-5.2, and GLM-5.3. The 2026 product cycle included open-source releases, larger context windows, improved coding performance, and changes to API pricing. The company also encountered compute shortages that affected access and prompted restrictions on new registrations. These developments illustrate Z.ai's evolution from a university-linked language-model startup into a publicly traded AI company with research, infrastructure, application, and international ambitions.

2026
Hong Kong listing
[1]Z.ai - Wikipediaen.wikipedia.org
Z.ai began trading on the Hong Kong Stock Exchange on 8 January.
2025
International rebrand and open-weight releases
[1]Z.ai - Wikipediaen.wikipedia.org
The company adopted Z.ai internationally and released selected GLM models under the MIT License.
2024
Expansion into video and speech generation
[1]Z.ai - Wikipediaen.wikipedia.org
The company announced Sora-like development work, released Ying, launched GLM-4-Plus, and introduced GLM-4-Voice.
2023
Major strategic financing
[1]Z.ai - Wikipediaen.wikipedia.org
Zhipu AI raised financing involving Alibaba Group, Tencent, Meituan, Ant Group, Xiaomi, and HongShan.
2022
GLM training method published
[1]Z.ai - Wikipediaen.wikipedia.org
Researchers presented the General Language Model training approach using autoregressive blank infilling.
2019
Company founded
[1]Z.ai - Wikipediaen.wikipedia.org
The startup emerged from research associated with Tsinghua University and was established as an independent company.

Products and positioning

Chinese AI technology company positioned around large language models, multimodal generative AI, open-weight model releases, developer infrastructure, and enterprise applications.

Large language model family · 2023
GLM
GLM, short for General Language Model and previously associated with ChatGLM, is Z.ai's principal family of pretrained and conversational language models. The series has expanded from dialogue systems into general-purpose, multimodal, speech, and coding-oriented models. Later releases include GLM-4.5, GLM-4.6, GLM-4.7, GLM-5, GLM-5.1, GLM-5.2, and GLM-5.3. Selected models were released with open weights under the MIT License.
Text-to-video model · 2024
Ying
Ying is a generative video system that converts textual or image-based prompts into short video clips. Its introduction marked Z.ai's expansion beyond text generation into audiovisual creation and Sora-style video research.
AI agent
AutoGLM
AutoGLM is an agent application designed to execute smartphone tasks from voice commands. It can interpret multi-step requests, interact with mobile services, and support activities such as ordering products or repeating purchases based on shopping history.
Research-information database · 2006
AMiner
AMiner is a scholarly and research-information database founded by Jie Tang in 2006. It provides structured information about academic publications, researchers, and research networks and is now owned by Z.ai.

Controversies

2025
Source

2026
Restricted new user registrations during compute shortageOther
Background
A shortage of computing resources reportedly caused service pressure and user complaints.
Action taken
Z.ai issued a public request for support and restricted new user signups.
Aftermath
The incident was accompanied by a reported decline in the company's share price.
Source
2026
Increased API pricesPrice change
Background
The company was expanding the GLM-5.1 platform while competing with premium United States model providers.
Action taken
Z.ai increased API prices by 10% while maintaining a lower price than Anthropic's cited Opus 4.6 model.
Aftermath
The announcement coincided with a reported rise in the company's share price.
Source
2025
Adopted Z.ai as the international brandStrategy
Background
The company had previously been known internationally as Zhipu AI while developing a global presence.
Action taken
It rebranded its international operations and model ecosystem under the Z.ai name.
Aftermath
The new identity was used alongside the company's Chinese corporate name and its GLM product family.
Source
2025
Released selected GLM models under the MIT LicenseStrategy
Background
Z.ai was competing for developers and users in a market dominated by both proprietary and openly released foundation models.
Action taken
The company made selected models available under the permissive MIT License.
Aftermath
The move strengthened Z.ai's open-weight positioning and broadened potential use by developers and enterprises.
Source

Recent events

Unconfirmed
Zhipu confirms developing Ox Alpha open-source model[8]智谱确认开发Ox Alpha开源模型chinaz.com
Zhipu confirmed that the anonymous open-source model Ox Alpha, which drew developer attention on OpenRouter, was developed by it and belongs to the latest GLM series. Zhipu said Ox Alpha will open its model weights on Wednesday, allowing developers to build on it; the model targets coding, continuous intelligent work and production-environment tasks, including long-term software engineering, complex reasoning and multimodal workflows combining text and visual context.
Weekdays and relative dates follow the original report; their exact calendar dates have not been verified.
Unconfirmed
Zhipu's next-generation GLM-5.4/5.5 large model to be released with trillion-parameter scale and long-horizon agent capabilities[9]智谱下一代大模型GLM-5.4/5.5即将发布：万亿参数、长程智能体能力ai.zol.com.cn
Reported: 2026-10-08
Zhipu's next-generation large model, expected by industry observers to be named GLM-5.4 or GLM-5.5, is set for imminent release with a significant leap over the current version. Founder Tang Jie, in a public speech in Singapore, outlined three R&D priorities: scaling pre-training data to the trillions of tokens with parameters expected to exceed one trillion (versus the current 740 billion); deepening post-training across supervised fine-tuning, RLHF and verification-feedback reinforcement learning; and substantively advancing long-horizon reasoning and agent capabilities. GLM-5.4/5.5 will be the first to integrate partial recursive self-improvement (RSI) capability components, while the complete 'fully self-training' vision points to a future GLM-6 generation.
2026-10-06
Amazon Bedrock integrates Zhipu GLM-5.3 with revenue-sharing cooperation[10]亚马逊云科技Amazon Bedrock正式接入智谱GLM-5.3并达成收入分成合作finance.sina.com.cn
Reported: 2026-10-08
On October 6, Amazon Web Services' large model platform Amazon Bedrock officially integrated Zhipu's GLM-5.3 model and announced it will share revenue with Zhipu based on model call volume. Zhipu has recently rolled out similar revenue-sharing cooperation models with multiple overseas cloud providers.
Unconfirmed
Two U.S. House Committees investigate the impact of rising adoption of Chinese AI models[11]美国众议院两个委员会调查中国AI模型使用率上升的影响cnbc.com
Reported: 2026-09-26
Two U.S. House Committees are investigating the impact of rising adoption of Chinese AI models, amid concerns in Washington about technology competition, security and Beijing's global influence. Chinese AI models from companies such as DeepSeek, Z.ai and Alibaba, which released new models with major performance gains, have grown to a majority share of token usage on developer platforms OpenRouter and Vercel in 2026. No date for the start of the investigation is given in the source.
Unconfirmed
Z.ai disables AI coding assistant features after security issue[12]Z.ai在安全问题发生后禁用AI编程助手功能news.google.com
Reported: 2026-09-21
Chinese AI company Z.ai disabled features of its AI coding assistant following a security issue.
Unconfirmed
Zhipu open-sources GLM-5.3 with commercial safety review threshold[13]智谱开源 GLM-5.3，百亿美元模型服务商商业使用须先过安全审查ifanr.com
Reported: 2026-08-29
Zhipu released the FP8 and BF16 weights of GLM-5.3 on Hugging Face. The model builds on GLM-5.2 with post-training improvements focused on complex coding, long-horizon tasks and vulnerability discovery, and its new license requires model-as-a-service providers with over $10 billion in combined 12-month revenue to pass a Zhipu security review before commercial use.
2026-08-27
Zhipu AI launches GLM-5.3 Flash lightweight flagship model[14]智谱AI发布轻量级旗舰模型GLM-5.3 Flashtech.ifeng.com
Reported: 2026-08-26
Zhipu AI officially launched GLM-5.3 Flash, a lightweight flagship model positioned against DeepSeek V4 Flash, featuring a 320B-parameter MoE architecture with 18B activated parameters, native multimodal input, and online inference served on a cluster of over 100,000 domestic chips.
2026-08-26
Zhipu releases and open-sources GLM-5.3-Flash multimodal model[15]智谱上线并开源GLM-5.3-Flash多模态模型citnews.com.cn
On the evening of August 26, Zhipu launched and open-sourced GLM-5.3-Flash (320B-A18B), the first native multimodal model in the GLM-5 series, and said Ox Alpha, which recently topped the OpenRouter usage ranking, belongs to the new GLM series. The model set new call-volume highs on OpenCode and OpenRouter, with all request traffic supported by domestic chips.
2026-08-26
Z.AI releases GLM-5.3-Flash large language model[16]Z.AI 推出 GLM-5.3-Flash 大型语言模型n.yam.com
Reported: 2026-08-27
Z.AI confirmed that its model codenamed Ox Alpha is the newest iteration of its GLM series, officially named GLM-5.3-Flash, and announced it would release the model's weights, allowing external developers to access its core parameters. The model runs on China-developed AI chips and is positioned as competitive with DeepSeek's leading models.
2026-08-17
Z.ai launches GLM-5.3 model update[17]Z.ai 发布 GLM-5.3 模型更新technode.com
Reported: 2026-08-17
Z.ai, the international brand of Chinese AI company Zhipu, launched GLM-5.3, an update focused on coding, long-horizon tasks and cybersecurity. The model uses the same base model as GLM-5.2, with gains attributed to post-training. Z.ai said GLM-5.3 scored 50% higher than GLM-5.2 on its internal Z.ai Code Bench and reached leading open-source results on Terminal-Bench 3.0 and Agents' Last Exam, and said the model weights will be released two weeks after launch following additional safety evaluation.

Showing the 10 most recent.

Sources37 sources

Cite this profile: Cite the canonical profile. /brand-wiki/zhipu-ai · Editorial policy · How profiles are compiled