DeepSeek has started rationing its free chatbot. Users of the Chinese AI lab's app and website reported on November 16 that long conversations, long prompts and even daily chats are now capped, with the company blaming a shortage of computing power.
The limits appeared without a formal announcement. Users discovered them through in-app notices and by hitting them in normal use, and DeepSeek has not said when — or whether — the restrictions will be lifted.
What changed for free users
The restrictions cover three things. Conversations left inactive for more than seven days can no longer be resumed, with users told that the chat is no longer supported "due to insufficient computing power." A single input is now capped at roughly 40,000 characters, narrowing the space for processing long reports, code files or batch material in one prompt. And a daily conversation quota now applies: users who hit it see a message saying the server is busy and their usage quota has been reached.
Some users reported a shorter inactivity threshold of 72 hours, and the exact size of the daily quota has not been published. The limits apply only to the free consumer app and website. DeepSeek's developer API, which runs on separate call limits and pay-as-you-go billing, is unaffected.
A capacity squeeze, not a bug
Industry analysts say the move is aimed at reclaiming server memory — known as KV cache — tied up by idle sessions. Demand has clearly outrun the computing capacity DeepSeek has available: its DeepSeek V4.1 Flash model handled 59.2 trillion tokens of calls in the week of November 9 to 15, ranking first globally, according to usage data cited in Chinese financial media.
It is not the first quiet cutback. In May, DeepSeek limited the regenerate and edit functions on its free chat without a public announcement. But a daily message cap is new: the company had never previously published one for its free service. The topic trended on Chinese social platform Weibo as users compared notes on the new limits.
Why it matters beyond DeepSeek
The rationing is a rare public glimpse of the chip squeeze that US export controls were designed to create. DeepSeek, China's best-known AI lab, built its reputation on doing more with less compute — and now says it does not have enough of it to keep its free service unrestricted.
The timing is pointed. The US Commerce Department has just moved to close a loophole that let Chinese firms rent advanced American AI chips through data centers abroad, a rule that directly targets the workaround for exactly this kind of shortage. DeepSeek's limits show the pressure is already real on the ground, whatever the workaround.
For now, the details rest on user reports and the notices inside the app itself. Confirmation from DeepSeek, or an explanation of whether the limits are temporary, will show whether this is a short-term capacity squeeze or a lasting constraint on one of the world's most-used free chatbots.



