DeepSeek Limits Long Conversations and Input Length, Users Call for a Paid Plan
Starting November 15, DeepSeek imposed limits on some features of its app and web version, covering long conversations, single-input character counts, and daily usage quotas; the official reason given was insufficient computing power. Some users reported that long conversations inactive for more than 7 days could not be continued, with some showing a 72-hour limit; when uploading a 430-page file, the system prompted that it could only read the first 20 pages or 10,000 tokens, and in testing it read only 10 pages. According to DeepSeek's official service status page, a total of 10 service incidents were recorded from October 1 to November 16, 8 of which involved the conversation service, and a large number of users in the comments section demanded the addition of a paid system.
Why read this
AI score 65
The rate-limiting rules corroborate the official service incident records, allowing readers to observe the trade-off free large-model services make between computing costs and user retention.
Topics
Body · AI translation
DeepSeek's long conversations restricted, netizens call out to Liang Wenfeng: charge us some money. Zhi Dong Xi download the client | Exclusive first look. November 16, 2026, 16:11:19, from Beijing
Zhi Dong Xi reported on November 16 that since yesterday, news has been appearing online one after another that DeepSeek has begun imposing usage restrictions on some functions of its App and web version, involving long conversations, the number of characters per single input, and daily usage quotas.
Some users reported that DeepSeek, citing "insufficient computing power," restricted long conversations that had been inactive for more than 7 days from continuing, with some users seeing a 72-hour limit.
In addition, some users encountered prompts such as "Content too long, please shorten and try again" and "Message count has reached the limit, please start a new conversation"—that is, DeepSeek's single-input text volume and number of conversation turns are now both restricted.
For example, after Zhi Dong Xi uploaded Anthropic's 430-page prospectus file, DeepSeek prompted that due to insufficient computing power, it could only read the first 20 pages or 10,000 tokens; in actual testing, it read only 10 pages.
Now, the comment sections of DeepSeek's official social media accounts have erupted into heated debate, with a large number of users complaining that these usage restrictions are unreasonable, and many users requesting that a paid system be added. Many users hope that by paying, they can continue using the long conversations they had accumulated previously.
Judging from the comments, among the users affected by the restrictions, many use DeepSeek for scenarios such as text games, interactive fiction, role-playing, and emotional companionship. These users have an extremely high emotional dependence on DeepSeek; some netizens even said that a conversation being interrupted felt like a "breakup."
They often keep advancing the plot in the same conversation window, accumulating a large amount of story settings and dialogue content. At the same time, these users play different games in different conversations, so even if they are still using DeepSeek regularly, they may not open the same conversation window within 7 days.
But not everyone in the comment sections supports DeepSeek launching a paid feature or rolling back the restrictions. Some users believe that DeepSeek itself is not a product specifically aimed at role-playing and emotional companionship, and that these scenarios should not be regarded as its main positioning.
Some users also said that the insufficient computing power problem they face cannot be solved simply by charging money.
In fact, the reason of insufficient computing power is not groundless; in reality, DeepSeek's service has been rather unstable recently.
According to DeepSeek's official service status history page, from October 1 to November 16, DeepSeek recorded a total of 10 service anomalies, 8 of which involved the conversation service.
Among them, on October 15, DeepSeek's web and API services experienced a partial outage lasting about 4 hours and 52 minutes.
Entering November, DeepSeek again experienced service anomalies on the 3rd, 7th, and 15th. Among them, the partial outage on November 3 lasted about 1 hour and 50 minutes, and the partial outage on November 7 lasted 45 minutes.
The anomaly closest to this rate-limiting incident occurred on November 15, when DeepSeek's conversation service experienced performance degradation and recovered after about 22 minutes.
It can be seen that since October, the frequency of DeepSeek service issues has been accelerating.
Just two days ago, a developer discovered a quota management mechanism in the front-end code of DeepSeek's web version, sparking speculation that DeepSeek might launch a personal paid plan.
It is worth noting that DeepSeek had just launched the DeepSeek full-suite application API on November 13, opening all the App's capabilities in the form of interfaces—that is, users can replicate the DeepSeek App themselves, with the only difference being that computing power is no longer free and an API must be connected.
As of press time, DeepSeek has not yet officially announced the complete rules regarding this rate limiting.
Conclusion: DeepSeek, short on computing power, and users with high emotional dependence are in conflict
This rate limiting has revealed another side of DeepSeek's users. In the past, when we talked about DeepSeek, it was often about technology, the frontier, and developers, but imperceptibly, DeepSeek's long text and excellent caching mechanism have led it to develop some "side businesses."
In addition to answering questions, writing code, and processing documents, some users also use DeepSeek for novel creation, interactive plots, and emotional companionship, and have built their own story worlds through long-term use.
Of course, users' willingness to pay does not mean that paying will necessarily solve the computing power and context management problems brought by long conversations.
For DeepSeek, how to balance free service, computing power costs, and the usage needs of different users is becoming an increasingly real problem.
"Special statement: The above works (including videos, pictures, and audio if any) are uploaded and published by users of 'Dafeng Hao,' a self-media platform under Phoenix Net. This platform only provides information storage space services.
Notice: The content above (including the videos, pictures and audios if any) is uploaded and posted by the user of Dafeng Hao, which is a social media platform and merely provides information storage space services."