DeepSeek Users Report Daily Chat Limits as Computing Shortage Strains Free Service

9 Min Read
DeepSeek Users Report Daily Chat Limits as Computing Shortage Strains Free Service

New limits reach DeepSeek's app and website

DeepSeek users reported new restrictions on November 16, 2026, including daily conversation quotas, a limit of about 40,000 Chinese characters per input and restrictions on long conversations left inactive for more than seven days. Notices displayed in the app and website cited insufficient computing power, bringing the pressure on China's artificial intelligence infrastructure directly into everyday conversations.

Contents
  1. New limits reach DeepSeek's app and website
  2. What users say has changed
  3. Why long conversations use more computing power
  4. From May restrictions to November quotas
  5. A possible alternative through the API
  6. How chip restrictions reach a chatbot
  7. Domestic chips offer a route with practical costs
  8. What the changes mean for users and suppliers
  9. Key Points

The daily allowance has not been disclosed. Users reaching it said they received a message stating that the service was busy and their usage quota had been reached. DeepSeek had not issued a public statement confirming the changes, explaining their duration or setting out how the allowances work.

The reported restrictions change the experience of a free service that had previously handled heavy demand through slowdowns and busy messages, without publishing a daily message cap. DeepSeek offers no confirmed paid consumer tier that allows users to skip the queue.

The changes also arrive shortly after DeepSeek opened additional tools through its application programming interface, or API, on November 13. That route lets software connect directly to DeepSeek's services and charges according to use. For people whose work depends on lengthy, continuing conversations, the developments create a practical choice between adapting to the free service's limits and using a route that carries a bill.

What users say has changed

The reported policy affects three parts of a conversation: how long an inactive thread remains supported, how much text a person can submit at once and how much they can chat in a day. Each restriction addresses a different part of the work required to operate an AI assistant.

Long conversations become restricted after more than seven days without activity, according to the notices described by users. Someone returning to an extended research or writing discussion after a break may therefore have to move that work into another conversation or service.

Advertisement

Single submissions are now reportedly limited to about 40,000 Chinese characters. An English language account described the figure as 40,000 words, while the Chinese wording and another detailed account specify characters. Characters and words measure different amounts of text, especially across languages, so the practical limit is best understood in the Chinese character terms reported by users.

The daily cap introduces another constraint even for people who send short messages. A person can reach a daily allowance without submitting a large document or maintaining a particularly long thread. The reported congestion notice combines a reference to busy servers with confirmation that the user's quota has been exhausted.

Why long conversations use more computing power

DeepSeek needs computing capacity each time it produces an answer. This stage is called inference: the model takes the user's request, processes the relevant conversation and generates a response. It is different from training, when a model learns patterns from large amounts of data before being made available.

A long conversation can require the system to process substantial amounts of earlier text so that an answer fits what has already been discussed. Large pasted documents add more material. Handling that work for many people at once puts pressure on processors, memory and the systems connecting them.

The reported restrictions address both the size and frequency of that work. A smaller input limit reduces how much text can arrive in a single request. A daily allowance limits repeated requests. Restricting inactive, lengthy conversations narrows the circumstances in which users can resume those threads.

DeepSeek's notices point to a shortage of computing capacity, but the company has not provided a technical breakdown of the constraint. The practical issue for users is immediate: access now appears to depend on a defined allowance as well as whether the service can handle demand at that moment.

From May restrictions to November quotas

DeepSeek had already shown signs of pressure before the latest reports. In late May, users found limits on how often they could regenerate an answer or edit a response. Those functions can trigger additional model processing, even when the user stays within the same conversation.

Advertisement

The November changes extend the reported restrictions beyond those individual functions. They affect the amount of text submitted, continued access to older conversations and the number of conversations a person can have during a day. The sequence of developments is:

  • Late May: Users found restrictions on regenerating and editing responses.
  • November 13: DeepSeek opened APIs for web search, file reading and voice features, and released a demo client that can import conversation histories.
  • November 16: Users reported limits affecting inactive long conversations, individual inputs and daily use in the app and website.

The sequence brings two different ways of accessing DeepSeek into focus. The free consumer service is being rationed, according to users, while the API provides access billed according to actual use. People choosing between them must weigh convenience, continuity and cost.

A possible alternative through the API

DeepSeek's November 13 release included an open source demonstration client that can import chat histories. Users can export their existing histories, bring them into that client and continue conversations through the API.

An API is a connection that allows one piece of software to request a service from another. The demonstration client supplies an interface for that connection, rather than requiring every user to build one. Usage is charged according to the amount consumed, making it a different arrangement from a free chat allowance.

For someone working through a long document or an ongoing project, importing a history offers a way to carry previous exchanges into another interface. The financial tradeoff is that continued use generates charges. The precise cost will depend on the relevant API pricing and how much the person uses it.

DeepSeek's web interface code also reportedly tracks the remaining quota for each model and refreshes that information every five minutes. That refresh describes how often the interface updates quota information; users still need a published policy to know when their allowance becomes available again.

A label reading "Coding Plan" in the interface code has prompted speculation about a paid service. No price or launch date has been confirmed. DeepSeek users currently have a reported free service cap and an existing API billing route, rather than an announced subscription that removes the consumer restrictions.

Advertisement

How chip restrictions reach a chatbot

US export controls restrict China's direct access to Nvidia's most advanced AI processors. These chips, often called accelerators, are designed to perform the many calculations that AI systems need at high speed. Access to enough of them helps determine how many requests a developer can serve and how quickly answers arrive.

Chinese companies have sought computing resources through routes including smuggled hardware and remote access to data centers in Southeast Asia. Remote access means the processors remain abroad while a customer sends computing tasks to them over a network.

The US Commerce Department is reportedly drafting rules that would restrict chip exports to Malaysia, Indonesia and Singapore and block Chinese remote access. The proposal would put pressure on a route developers have used to obtain computing capacity beyond China's borders. Its final terms and implementation schedule have not been announced.

Taiwan is also seeking tighter controls on physical shipments. In October, Economic Affairs Minister Kung Ming-hsin described plans for criminal penalties for chip smuggling, alongside the addition of advanced AI chips and servers to controlled commodity lists. That proposal would give prosecutors a charge directed at unauthorized exports to China, with no implementation date specified.

Domestic chips offer a route with practical costs

China has been directing publicly supported computing projects toward domestic processors. Guidance described in November 2025 required new data centers receiving state funding to use domestically produced AI chips. Projects less than 30 percent complete were told to remove installed foreign accelerators or cancel planned purchases, while more advanced projects faced individual assessments.

The domestic chip requirement reached a construction program that had attracted more than $100 billion in public funding since 2021. Support could include grants, tax incentives, discounted land, electricity assistance or government equity, bringing several forms of public backing within the reported policy.

Domestic suppliers positioned to serve that demand include Huawei, Cambricon, Enflame, MetaX, Moore Threads and Biren. For AI developers, however, replacing processors involves more than selecting a different brand. Servers, cooling, networking and the software that runs models may also need changes.

Advertisement

DeepSeek's reported restrictions therefore arrive during a transition in the supply of computing equipment. Restrictions on advanced foreign chips increase the incentive to buy domestic alternatives, while integrating those alternatives takes resources and engineering work. Users experience the result at the simplest level: whether a conversation continues when they need it.

What the changes mean for users and suppliers

People using DeepSeek for repeated drafting, coding or document analysis face the clearest practical disruption. A daily cap can interrupt a session before a task is finished, while the inactivity rule affects projects resumed after a week away. The reported input limit also constrains how much material can be submitted in one request.

Domestic chipmakers and AI infrastructure suppliers may benefit from pressure to expand local computing capacity. The scale of that opportunity depends on developers' purchasing decisions, the equipment suppliers can deliver and how readily that hardware supports the required software.

DeepSeek's next product or capacity announcement would help users plan their work. A published quota, a statement on additional computing resources or an announced paid service would each provide concrete information about access. For now, users are encountering the restrictions through the app and website while the company has yet to explain the policy publicly.

Key Points

  • Users reported new limits across DeepSeek's app and website on November 16, 2026.
  • The restrictions cover long conversations inactive for more than seven days, inputs of about 40,000 Chinese characters and an undisclosed daily allowance.
  • DeepSeek has not publicly confirmed the policy or said how long it will last.
  • An API demo client can import chat histories, with API use billed according to consumption.
  • US chip restrictions and China's shift toward domestic processors form the wider computing backdrop.
Share This Article

You May also Like