DeepSeek recently imposed limits on long conversations for its app and web version, covering multiple aspects such as the retention of conversation history, the maximum length of a single input, and daily conversation quotas. According to user feedback, this adjustment is mainly related to the platform's current pressure on compute resource capacity.

QQ20261116-141859.jpg

Specifically, DeepSeek currently no longer supports long conversation histories that have been inactive for more than 7 days, and the text input limit for a single question has also been lowered to about 40,000 characters. At the same time, the platform has added a daily conversation quota limit. When a user's usage reaches the limit, the system will display a message saying, "Due to server congestion, the usage quota has been reached," temporarily preventing further use.

This adjustment reflects the challenge large model products face in balancing compute supply and service demand against the backdrop of continuously growing user scale and usage demand. As the daily active users and interaction frequency of AI applications continue to rise, features such as long-context conversations and multi-turn interactions will also continue to consume computing resources.

How to control operating costs while ensuring user experience through resource scheduling, inference efficiency optimization, and service tiering has become an important issue that AI platforms need to address.