AI giant Zhipu apologizes and announces handling and compensation plan

Zhipu issued an apology letter on February 21 regarding the GLM Coding Plan and announced a handling and compensation plan.

Zhipu stated that the main mistakes in this revision were: insufficient transparency of rules, slow rollout of the GLM-5 grayscale phase, and rough upgrade mechanisms for old users.

Zhipu explained that after the release of GLM-5, traffic exceeded expectations, and the company’s expansion pace could not keep up. As a result, GLM-5 was gradually opened in the order of Max, Pro, and Lite. Currently, Max users have been fully granted access, Pro users have been opened but may experience rate limiting during peak times due to high cluster load, and Lite users will be gradually opened in non-peak periods after the holiday. For affected Lite and Pro users, the company supports self-initiated refund applications.

It is noteworthy that on February 20, AI large model leader Zhipu (02513.HK) stock price rose against the trend, surging 42.72% in a single day to HKD 725, reaching a new high, with a total market value of HKD 323.2 billion. In just 43 days since listing, the company’s stock price has increased by 524% from the issuance price of HKD 116.2.

On February 20, Kuaishou (01024.HK) closed down 2.78%, with a total market value of HKD 289.4 billion; JD.com (09618.HK) fell nearly 2%, with a market value of HKD 330.9 billion; Ctrip (09961.HK) declined nearly 1%, with a market value of HKD 296.7 billion; Baidu Group (09888.HK) dropped 6.25%, with a market value of HKD 354.8 billion. This indicates that Zhipu has entered the top tier of market capitalization in the Hong Kong stock TMT sector.

Zhipu officially launched its new flagship model GLM-5 on February 12, which has an average performance improvement of over 20% in programming development scenarios compared to the previous generation, with a real programming experience approaching Claude Opus 4.5 level. It achieved the best performance in three agent evaluations: BrowseComp, MCP-Atlas, and τ2-Bench, among open-source projects.

After the release of GLM-5, due to high demand, Zhipu immediately increased the prices of the GLM Coding Plan packages the next day. The price in China increased by 30%, and the overseas version increased by over 100%, making it the first domestic AI native company to raise prices for large model commercialization services. The new packages sold out immediately, setting a new industry record for paid Chinese AI programming model packages.

Founded in 2019, Zhipu originated from Tsinghua University’s Knowledge Engineering Laboratory (KEG, one of China’s leading AI laboratories). Professor Tang Jie from Tsinghua’s Computer Science Department serves as Chief Scientist, Dr. Zhang Peng, a leading engineer at Tsinghua, is CEO and General Manager, and Tsinghua alumnus Liu Debing is Chairman.

This company is known as “China’s OpenAI” and is the top competitor of OpenAI. At the end of 2020, Zhipu developed the GLM pre-training architecture; in 2021, it trained the 10-billion-parameter GLM-10B model, and the same year successfully trained a converged trillion-sparse model using MoE architecture; in 2022, it released and open-sourced the Chinese-English bilingual 130-billion-scale pre-trained model GLM-130B, which was the only Asian large model selected for Stanford evaluation in 2022.

Zhipu’s full apology statement:

(Source: Daily Economic News)