According to AIbase, leading domestic large model company Zhipu AI is facing a "happy problem". On January 21, Zhipu announced on its official WeChat account that due to a surge in user numbers after the launch of GLM-4.7, computing resources have become temporarily scarce. To ensure the experience of existing users, Zhipu has decided to implement limited sales of GLM Coding Plan, actively reducing the load on servers.

QQ20260121-143818.png

Computing Power Crisis: Peak Time Concurrency Limiting

With the comprehensive upgrade of GLM-4.7's performance, the number of subscriptions for GLM Coding Plan has seen explosive growth. According to Zhipu's observations, some users recently encountered frequent concurrency limiting errors and slower model responses during peak usage times on weekdays between 15:00-18:00. This phenomenon indicates that the current surge in programming computing demand has approached the capacity limit of Zhipu's existing computing cluster.

Limit Sales Upgrade: Daily Quota Reduced to 20%

To optimize the programming experience for existing users, Zhipu has decided to take extreme "supply protection" measures. Starting at 10:00 on January 23, GLM Coding Plan will enter a limited sales phase, with the daily available quota significantly reduced to 20% of the current level. The quota will be refreshed at 10:00 every morning. Notably, users who have activated automatic renewal will not be affected and can continue to use it normally by deducting fees.

Response Strategy: Trade "Quantity" for "Quality"

Zhipu stated that this move aims to prioritize valuable computing resources for existing subscribers. Although reducing the daily new sign-up quota by 80% may sacrifice some market growth, it can effectively ease congestion during peak hours and ensure smooth operation in core programming scenarios. The specific time for stopping the limit sales and restoring full supply will be announced separately by Zhipu based on the progress of computing power expansion.