DeepSeek free limits: 40,000-character cap and daily quota
DeepSeek has capped free chats at 40,000 characters and added a daily message quota. Demand for its V4.1-Flash-1017 model hit 59.2 trillion tokens in a week.

If DeepSeek's free chatbot has started telling you the server is busy, it isn't just you. The company has tightened the DeepSeek free limits in its app and web version, and it blames a shortage of computing power.
Three changes stand out, based on user reports. Long conversations that have sat idle for more than seven days can no longer be continued (some users got a 72-hour warning first). A single request is now capped at roughly 40,000 characters. And there is a daily message quota: when it runs out, you see "server busy, usage quota exhausted". Some users say they get cut off after just two hours of continuous use. The company hasn't explained the quota rules or said when the limits will lift.
Why is DeepSeek limiting free users?
Demand has outrun capacity. On the API aggregator OpenRouter, the V4.1-Flash-1017 model, updated on 17 October, processed 33.6 trillion tokens in the week of 19 to 25 October, up 801% on the week before. By 9 to 15 November that had grown to 59.2 trillion tokens, enough to keep it first worldwide. The model launched on 10 September handled 15.8 trillion in its comparable week, so volume nearly tripled in two months.
The service has struggled. DeepSeek's status page lists at least eight incidents since October, including a more than four-hour outage on 15 October that hit both the chat interface and the API.
Does this affect developers using the API?
No. The limits apply to the consumer app and website. The API is pay-as-you-go with separate concurrency caps: 2,500 sessions for deepseek-flash and 500 for v4-pro. Off-peak API prices are half the usual rate. On 13 November DeepSeek also opened a full API covering web search, deep thinking, files and links, and voice input, and released its chat interface code under the MIT licence.
The hardest hit are those who used the free tier for very long contexts, such as big documents or large chunks of code. Code in the web interface had sparked rumours of a paid subscription, but it contained no prices, sign-up pages or payment tools.
Help is some way off. DeepSeek plans a gigawatt-scale data centre in Ulanqab with at least 160,000 Chinese-made AI chips, coming online between late 2027 and early 2028.
Apps worth a look
Tap the right tiles, add them up, hit the target. Easy to learn, hard to master. Free, for all ages.
Talk to an AI conversation partner; every sentence comes back corrected and explained in your language. 9 languages, A1 to C2.
Open EDIFACT and X12 messages, see every segment explained, convert to JSON/XML, export PDF and Excel. Runs locally, no upload.
Pick who it is for, the occasion and their personality. The search engine reads product descriptions from local shops and ranks what fits. Free, no sign-up.
More from this day
See all →
AI chatbot trackers leak your chats: 21 of 42 services tested

floppy3.5: a Wi-Fi floppy reader for old Sony Mavica photos

Wii U vWii color bleed traced to two typos in 13-year-old firmware




