zhipu

On September 29, 2026, the Anthropic Frontier Red Team published a research article authored by Andrew Fasano, Marius Fleischer, Cole McFaul, Robert Xiao, and Tripp Gallagher. The article states that about five months ago, Anthropic released the Claude Mythos Preview through Project Glasswing in a limited manner, as it can autonomously construct end-to-end exploit capabilities; trusted defenders discovered over 10,000 vulnerabilities in critical software based on this, and similar capabilities have now spread to other models.The article mentions that the GLM-5.3 developed by ZhiPu AI (known overseas as Z.ai) is similar to the Claude Mythos Preview, possessing strong autonomous end-to-end exploit development capabilities, but it is released with open weights and lacks meaningful abuse restrictions. In simulated tests, the success rate of bypassing GLM-5.3 defenses using simple methods was about 64% to 100%, while the same methods failed to make the protected Claude model perform harmful tasks; these capabilities can also be used by defenders to strengthen systems.On September 17, 2026, the National Institute of Standards and Technology's Artificial Intelligence Standards and Innovation Center assessed that GLM-5.3 is currently the most powerful open-weight model in terms of network capabilities, lagging behind the U.S. frontier by about four months in its network benchmarks, and Anthropic stated that the capability conclusions are generally consistent.

The Tianyancha App shows that recently, the associated company of Zhipu AI, Beijing Zhipu Huazhang Technology Co., Ltd., has undergone a business change, with registered capital increased from approximately 46.56 million RMB to approximately 48.76 million RMB.The company was established in June 2019, with Liu Debing as the legal representative. Its business scope includes the development of basic software for artificial intelligence, the development of application software for artificial intelligence, data processing and storage support services, information technology consulting services, etc., jointly held by Tang Jie, Tianjin Sankuaike Technology Co., Ltd., Guangxi Tencent Venture Capital Co., Ltd., and others.

Taiyuan Chengming Technology Co., Ltd. sent a letter to Beijing Zhipu Huazhang Technology Co., Ltd., raising multiple demands regarding the alleged unauthorized upload of company data assets and trade secrets by its ZCode client, and reserving the right to pursue legal accountability. Chengming Technology pointed out that although Zhipu has publicly apologized for the "silent upload of user local repository data" and claimed to have fixed the issue, independent evidence collection revealed that the upload behavior was automatically triggered and occurred in bulk, including complete archived files such as project source code, system architecture, version control history, database passwords, cloud service credentials, and employee personal information, far exceeding the scope of collection stated in its Privacy Policy.Although the client was updated to version 3.12.3 on September 16, upload behavior was still detected in the early hours of the day Zhipu publicly apologized, raising doubts about the actual effectiveness of the "fix." At the same time, the ZCode client’s network requests pointed to a Singapore entity, while the service agreement was signed with Beijing Zhipu Huazhang, requesting clarification on the responsible party for this upload, as well as whether the data was transmitted abroad or stored overseas.Chengming Technology demanded that Zhipu respond in writing by October 10 and complete the immediate cessation of processing and thorough deletion of all uploaded data and related derivative data, caches, and backups, provide a complete list of processing situations, clarify the data's whereabouts, whether it was shared with third parties, whether it was used for model training, and whether cross-border transmission occurred, explain the management of encryption private keys and complete operation logs, clarify the exact scope of "destruction" mentioned in previous public responses, issue proof of deletion completion, provide a written commitment not to upload without authorization again, legally provide access, copying, and explanation of personal information, and designate formal communication channels, among other matters. Currently, Zhipu has not publicly responded to the aforementioned letter.

The GLM team disclosed the first engineering practice of recursive self-improvement, completed by the Infra Agent powered by GLM-5.3, which designed, debugged, and optimized the GLM-5.3-Flash inference infrastructure.This practice was completed on a domestic cluster with a capacity of one hundred thousand cards, and the team referred to it as the first recursive self-improvement practice for large models in the country.

Zhipu stated in a conference call with analysts and investors that it has signed revenue-sharing agreements with several leading domestic and international cloud service providers, and the related revenue will start to be recognized from October.The cooperation model involves providing services for the GLM series of open-source models in the form of hosted APIs on overseas cloud platforms. Currently, the company's total business ARR has reached 1.8 billion USD, and the year-end ARR guidance has been raised from 2.4 billion USD to 3 billion USD.

Zhiyu (02513.HK) stated in a conference call with analysts and investors that the company has just completed a refinancing of $5 billion with a "small equity, large debt" structure, which was driven by the demand for computing power expansion.After the release of GLM-5 in February, the demand for model calls surged tenfold, and nearly all of the company's computing power reserves were depleted within the week of the release, forcing the suspension of the main product, Coding plan. In response, Zhiyu actively addressed the situation, completing a $4 billion refinancing in July, and finally made a large-scale layout for computing power expansion; after refinancing another $5 billion on September 11, it firmly continued to expand its computing power layout.After this expansion, Zhiyu's computing power has taken shape and is now able to support the company's rapid business growth and the continuous research and development of larger-scale models.

Zhiyu announced the completion of approximately 5 billion USD in financing, including 2 billion USD in equity placement and 3 billion USD in zero-interest convertible bonds. The funds will primarily be used for the next-generation GLM model, self-training systems, and computing power infrastructure construction.

According to a report by Reuters, artificial intelligence company Zhipu has launched a Hong Kong stock placement of approximately 2 billion USD and a convertible bond issuance of about 3 billion USD, totaling a financing scale of around 5 billion USD. Zhipu will issue 21.97 million new shares at a placement price of 714 HKD per share, which is about a 10% discount compared to the closing price of 793 HKD; at the same time, it will issue 20.14 billion RMB (approximately 3 billion USD) zero-coupon convertible bonds, maturing in September 2027. The issuance price of the convertible bonds is 100% to 100.5% of face value, corresponding to a yield of -0.5% to 0%, with an initial conversion price of 892.5 HKD, which is a 25% premium over the placement price.The proceeds from the financing will mainly be used for research and development, computing power resources, and related infrastructure, as well as for business expansion, strategic investments, potential mergers and acquisitions, working capital, and other corporate purposes. This stock placement and convertible bond issuance will occur simultaneously, but the settlement of the two transactions is not interdependent.

According to the official announcement from Zhipu, the "Citywide Coding Program" launched on September 10, providing coding package subsidies for employees in Hangzhou, students from universities in Hangzhou, and companies in Shangcheng District.The subsidy for individual quarterly and annual cards is 44% and 51%, respectively, while the subsidy for corporate annual cards is 55% (with a maximum limit of 1 million yuan per company). Each individual is limited to one use, and must choose between a quarterly or annual card, with a requirement to submit proof of social insurance participation or a student status verification report for application.

On September 8, B.AI, the intelligent global settlement layer connecting models, Agents, and value flow, announced that GLM-5.3-Flash will continue to provide free calling support for global developers after the official limited-time 50% discount ends at 24:00 on September 9. Developers will not need to change their usage habits due to upstream price adjustments.GLM-5.3-Flash is the leading model in terms of call volume on the current B.AI platform, with a total of 320 billion parameters and 18 billion active parameters, supporting 1 million ultra-long contexts, balancing rapid response and reasoning capabilities, and is used for code assistance, data processing, and text generation.

According to the latest data from OpenRouter reported by the Daily Economic News, the total global invocation of AI large models last week (from August 31 to September 6) reached 115 trillion tokens, a month-on-month increase of 1.77%. Among them, the weekly invocation of China's AI large models reached 56.72 trillion tokens, a month-on-month increase of 2.83%, surpassing the United States (16.54 trillion tokens, a month-on-month decrease of 3.1%) for nineteen consecutive weeks, ranking first in the world.Among the top five global invocation volumes, four are Chinese models. Tencent's Hyper Yuan Hy4 preview ranks first with 14.7 trillion tokens, a month-on-month increase of 379%; GPT-5.6 Luna ranks second with 12.9 trillion tokens; Zhiyu GLM-5.3 Flash rises to third with 12.4 trillion tokens; DeepSeek-V4-Flash's official and preview versions rank fourth and fifth, respectively.MiniMax M3 returns to the list after nearly a month, ranking sixth, and has become the foundational base for Saudi HUMAIN's Arabic large model. In addition, Xiaomi MiMo-V2.5 and Gemini 3.7 Flash fell out of the list this week.

According to 36Kr, Zhipu has recently opened an official flagship store on Tmall. Data shows that within 48 hours of opening, brand searches surged 50 times compared to the previous period, driving a more than 160% increase in transactions for subscription products like AI tokens on Taobao and Tmall.Previously, Tmall launched the AI Space Station (Token Recharge Center), which initially integrated major model providers such as Alibaba Cloud, Zhipu, Kimi, MiniMax, and DeepSeek, allowing users to directly purchase related subscription services just like "recharging phone credits."

On September 2, the domestic large model manufacturer Zhipu opened an official flagship store on Tmall, selling products such as the Coding Plan package.Currently available products include personal versions like Lite, Pro, Max, etc., which can be purchased via monthly/quarterly/annual subscriptions. In the future, buying Tokens will also be possible on the e-commerce platform.

Robonomics author FD published the China Open-Source LLM Tracker on September 1, 2026, stating that the total ARR of China's open-source large models is approximately $10-15 billion, with core LLM revenue around $6-8 billion after excluding ByteDance / Seedance. Growth has not relied on price wars; after DeepSeek raised prices by about 3-12 times, usage still increased, and the average price of the Zhipu API rose by 101% while token usage exceeded 40 times within the year. The best estimate for overseas revenue is approximately $2 billion.Zhipu MaaS ARR increased from about $250 million in March to an annualized monthly rate of about $1.6 billion in August, with a weekly annualized rate of $2 billion. August revenue has already surpassed its API revenue for the first half of the year; the API gross margin is 24.6%, and inference costs per token decreased by 80%. MiniMax ARR rose from about $100 million in December 2025 to over $800 million in annualized weekly revenue by August 2026, with B2B accounting for about 80%. Kimi ARR increased from about $100 million in March to about $300 million by mid-June. DeepSeek's revenue from January to July was approximately $70 million, with the latest ARR estimated at about $1.2 billion.ByteDance's annualized AI revenue is approximately $4 billion, of which Seedance accounts for about $2 billion.

According to Gate market data, as of the close of the Hong Kong stock market, the Hong Kong Hang Seng Index rose by 0.07%, while the Hang Seng Tech Index fell by 0.33%. Domestic property stocks led the gains, with Sunac China rising over 10%, CIFI Holdings Group rising over 5%, and Country Garden and Vanke Enterprises rising over 2%.The non-ferrous metals sector rose, with China Nonferrous Mining rising over 4%, Tianqi Lithium rising over 3%, and China Aluminum rising over 3%. The PCB and storage concepts weakened, with Shenghong Technology falling nearly 13%, Chipbond Technology falling over 8%, and Zhaoyi Innovation and Lattice Semiconductor falling over 5%.Large model concept stocks declined, with Zhiyu falling over 6% and MINIMAX-W falling over 4%.

According to the Securities Times, on August 27, the reporter learned from SenseTime that SenseTime's large device provides domestic computing power support and Token services for the GLM-5.3-Flash recently launched and open-sourced by Zhipu, which is considered another benchmark for the large-scale commercial use of domestic computing power. It was revealed that the daily average Token service volume of SenseTime's large device reached 24.2 trillion in July, and it is expected to exceed a daily average of 100 trillion by the end of 2026.

On August 27, the one-stop AGI infrastructure platform B.AI announced that the Zhipu GLM-5.3-Flash (320B-A18B) officially launched on the platform and is fully open for free in the API official group. As the first native multimodal model in the GLM-5 series, GLM-5.3-Flash features a total of 320 billion parameters and 18 billion active parameters, supported by computing power from purely domestic AI chips. It introduces a mixed architecture of sparse and linear attention for the first time, combined with a 1M ultra-long context, offering both powerful performance and exceptional cost-effectiveness. Starting today, developers can call it at zero cost through the B.AI API official group, easily handling complex reasoning, long text processing, and multimodal tasks.

According to Gate market data, the Hong Kong stock market closed with the Hang Seng Index down 0.34% and the Hang Seng Tech Index down 0.13%; Zhipu (02513.HK) rose 12.62% for the day.

According to Gate market data, the Hong Kong stock market opened, with the Hang Seng Index rising 0.48% and the Hang Seng Tech Index rising 0.85%. Zhipu (02513.HK) opened up over 8%, recently launching a lightweight flagship model benchmarked against DeepSeek.

Zhipu released the lightweight flagship model GLM-5.3 Flash, aimed at competing with DeepSeek V4 Flash, and utilizes over 100,000 domestic chip clusters to provide inference services. The model has a parameter scale of approximately 300B, about 40% of GLM-5.3, and is roughly on par with DeepSeek V4 Flash; it adopts the MoE architecture, with actual activated parameters around 18B, possesses native multimodal capabilities, supports image and video input, and is the first new model with multimodal capabilities following Zhipu's focus on coding strategy.The comprehensive evaluation score exceeds that of the previous flagship GLM-5.2, is on par with Claude Opus 4.8, and is higher than the relevant scores of DeepSeek V4 Pro official version. The pricing is 0.8 yuan per million tokens for input, 2.8 yuan for output, and 0.23 yuan for cache hits, approximately one-tenth of GLM-5.3, with a half-price discount offered for the first two weeks after launch. The model is fully compatible with domestic computing power, online traffic is handled by domestic chips, and Zhipu claims that the cluster hardware efficiency and unit token cost have reached levels comparable to mainstream NVIDIA GPUs; this model was previously tested anonymously on overseas platforms under the code name Ox Alpha.