Zhipu Restricts GLM Coding Plan Sales Due to Resource Shortage

Zhipu, a Chinese company developing large language models (LLM), has announced a temporary restriction on sales of its GLM Coding Plan product. This decision was made due to increased computational load on the company's infrastructure following the recent launch of a new version of their flagship model – GLM-4.7.

GLM Coding Plan is a service that provides developers with access to GLM capabilities for programming-related tasks. Since the launch of GLM-4.7, there has been a significant increase in the number of users, which has led to an overload of computational resources, particularly during peak hours (15:00 to 18:00 on weekdays).

During this time, some users have experienced errors related to concurrent request limitations as well as model performance slowdowns. Zhipu apologized to its users for the inconvenience and announced the beginning of work to expand its computational capacity. However, to ensure stable service operation for existing clients, the company decided to temporarily restrict sales of GLM Coding Plan.

Daily sales volume will be reduced by 80% from current levels. This decision, while temporary, highlights the growing problem of computational resource shortage for training and using large language models. As models become increasingly complex and powerful, the need for computational capacity increases significantly.

This creates challenges for LLM-developing companies in terms of infrastructure scaling and cost management. The restriction on GLM Coding Plan sales may affect developers using GLM for their projects. This could lead to delays in development or the need to find alternative solutions.

However, on the other hand, this could also encourage developers to use available resources more efficiently and to optimize their requests to the model. Overall, the GLM Coding Plan situation demonstrates growing competition for computational resources in the field of artificial intelligence. Companies working with LLMs will need to invest in expanding their infrastructure and in developing more efficient algorithms to meet the growing demand for their products and services.

In the future, it is expected that similar restrictions will emerge at other LLM-developing companies until more efficient solutions for scaling computational resources are found. The restriction on GLM Coding Plan sales is a temporary measure aimed at ensuring stable service operation for existing users. However, this situation is also a signal of the growing problem of computational resource shortage in the AI field and the need for long-term solutions for infrastructure scaling.