On the afternoon of May 29, some internet users noticed that DeepSeek has started to set limits on how many times users can regenerate and modify the responses they receive. In normal conversational mode, the limit is set between 3 to 6 regenerations. In expert mode, users are restricted to just 3 attempts, while the maximum number of times they can modify their inputs is typically 6. As of now, DeepSeek has not released an official statement or a detailed quota table outlining these restrictions. However, this action has already triggered discussions among its user base. In reality, when large language model products encounter challenges such as a shortage of computing resources, a sudden increase in user numbers, or excessive pressure during peak usage times, they often resort to measures like traffic throttling and service degradation to ease the burden on their servers. Other AI products, such as ChatGPT and Claude, have also adopted similar strategies.