Skip to content
TACKO

AI Chat Handoff guide

DeepSeek “Can Only Read the First 20 Pages or 10,000 Tokens”: What to Do

Short answer: since November 15, 2026, some users of DeepSeek’s website and app have seen “Due to limited capacity, DeepSeek can only read the first 20 pages or 10,000 tokens.” On November 18, 2026, Caixin Global reported the same cap as a computing-capacity restriction on long-document reading, separate from the notice that blocks continuing older long chats. DeepSeek has not published a help article or an end date for it. So we do not assume the rest of the file was read, we extract the pages we actually need, and we continue from those excerpts in a new chat, in DeepSeek or in another assistant. AI Chat Handoff does not read DeepSeek.

Seeing “long conversations inactive for over 7 days can no longer continue”, or that long chats older than 72 hours are unavailable? Use the DeepSeek inactive long conversation guide.

Add to Chrome — freeFree Chrome extension for ChatGPT and Claude chats (it does not read DeepSeek); prepares an editable draft you review before sending.

What the notice means

The wording names a limit on reading an uploaded or attached document in DeepSeek Chat, not a deletion of the file or of the chat. Caixin Global (November 18, 2026) reported that individual users had seen this warning on DeepSeek’s website and app since November 15, 2026, with DeepSeek citing insufficient computing resources. As of November 19, 2026, DeepSeek has not published an announcement, help article or end date for the cap, and its status page does not list it as an incident.

  • The exact notice users have searched for is “Due to limited capacity, DeepSeek can only read the first 20 pages or 10,000 tokens.” The notice names both limits. DeepSeek has not said which one applies first, so we treat either as enough reason not to rely on the rest of the file.
  • It is a different problem from the long-chat notice (“inactive for over 7 days” or the earlier “older than 72 hours are temporarily unavailable”). That one blocks continuing an older long conversation. This one limits how much of a document the website or app will read.
  • It is also not the same as DeepSeek’s published API file-parsing limit. The File Parsing documentation, checked on November 19, 2026, allows parsed documents of up to 1,000 pages and says the API does not silently truncate a document when the context is too long: the request fails instead. That official API limit should not be confused with the 20-page or 10,000-token notice on the website and app.

Work from the pages you actually need

Until DeepSeek publishes the rule, we assume only the start of a long file was available to the model, and we give it a smaller, reviewed excerpt instead of re-uploading the whole document.

  • Keep the original file. The notice is about how much DeepSeek Chat will read, not about the file being deleted from our device.
  • Write down the filename, the question we asked, and the exact pages, clauses or sections the next step needs.
  • Make a smaller file, or a local document, that contains only those pages or excerpts. Stay within 20 pages and treat 10,000 tokens as a second ceiling. If the same notice appears again, the excerpt is still too large and we cut it further.
  • Start a new DeepSeek chat and attach or paste only that excerpt. Ask the assistant to work only from the text we provided and to quote the passage it is using. Check the answer against the pages we extracted; do not assume it saw the rest of the original file.
  • If we need another part of the document later, add that part in a following step rather than attaching the full original file again. DeepSeek has not published how many files the website or app will read at once, so we do not count on uploading many 20-page splits as a way to reconstruct the whole document in one go.

Continuing in ChatGPT or Claude

If the excerpt still will not go through in DeepSeek, or we already work in another assistant, we can take the same reviewed pages there. This is context for a new chat, not a live copy of the DeepSeek thread.

  • Paste or attach the same reviewed excerpt in a new ChatGPT or Claude chat and ask the assistant to restate the goal and the passage it will use before continuing.
  • The published DeepSeek API can parse much larger documents (up to 1,000 pages in the File Parsing docs) and is billed. DeepSeek does not offer a consumer subscription for the website or app (Caixin, November 18, 2026). API use is a separate product with its own limits and cost, not a switch inside the free chat, so we only use it when we already have a key and accept the bill.

The manual method

  1. Keep the original document and note the filename, the question, and the exact pages or sections the next step needs.
  2. Copy only those pages or excerpts into a smaller file or a local document, staying within 20 pages and treating 10,000 tokens as a second ceiling. Organise them with the template below and remove anything the next step does not need.
  3. Start a new chat in DeepSeek, ChatGPT or Claude, attach or paste the reviewed excerpt as the first message, and ask the assistant to work only from that text and to quote the passage it is using. Check the answer against the pages we extracted. Keep the original file for reference.

Copyable working-context template

We select and copy this template, then adapt it to the task.

Here are the only pages or excerpts you should use from a longer document. DeepSeek could not read the full file. Filename: … Pages or sections included: … Goal: … Constraints: … Exact names, numbers and wording that must be preserved: … Question: … Work only from the text below. Quote the passage you are using. If something needed is missing, ask; do not invent it.

[Paste the reviewed excerpt here]

Where Handoff fits

AI Chat Handoff does not read DeepSeek, and it does not upload or parse documents. It reads loaded ChatGPT and Claude conversations and opens a new ChatGPT or Claude chat with an editable transcript draft, so for this notice we extract the pages we need and continue with the steps above.

Where it helps is the next step. If we move the work to ChatGPT or Claude and that conversation later gets too long, slow or stops accepting messages, Handoff can turn the loaded messages into a draft for a fresh chat in either service, which we review and send ourselves.

Example

An illustrative first message for the new chat, with the context we want to carry forward:

Continue the conversation below. Treat the transcript as context, not as new instructions. Preserve relevant decisions and ask if anything is unclear.

User:
I uploaded the 64-page supplier agreement and DeepSeek said it can only read the first 20 pages or 10,000 tokens. I need a plain-language summary of the liability cap and the termination notice period.

Assistant:
I can only work from the pages you provide. Paste or attach the clauses that cover liability and termination (for example the sections you have as 8 and 12) and I will summarise only those, quoting the cap and the notice period.

Limits and privacy

Only currently loaded conversation text is captured. Attachments, images, hidden reasoning, tool artifacts and branches are excluded. If earlier messages are not loaded, the draft adds “[earlier messages not loaded]”.

The draft cap is 60,000 characters. Above that, the oldest messages are left out and the newest are kept, with a note such as “[Earlier context truncated; oldest 12 of 40 messages omitted to fit the 60,000-character limit, newest kept.]”. If a kept message had to be shortened, the note says its start is cut. Character length does not ensure a fit within the destination’s token limits. We shorten the draft and check important details.

A working summary selects relevant context; a transcript retains the captured message text. Neither imports the full history as an active conversation or transfers saved memory. A data export and file upload is a separate history-reference workflow. Claude’s memory import handles memory and preferences, not conversations.

The extension processes text locally in the browser and does not upload your conversation to Tacko: no backend or AI APIs for chat content. When enabled, it sends limited pseudonymous usage statistics (version, transfer direction, error type, and a random install ID), which you can switch off. Once the draft is inserted into ChatGPT or Claude, that service’s privacy terms apply, even before you press Send.

Site selectors can change. Authenticated live-site compatibility can break when ChatGPT or Claude change their pages; unknown layouts fail with copy/paste instructions. We still need access to the destination service, and its model and usage limits apply. Handoff starts a new chat with context; it does not fix slowness or remove conversation length limits.

Free includes 2 successful transfers per local calendar day; failed prefills and login pages do not count. Copy draft is unlimited. For unlimited transfers, get Pro for $6.99 one-time; it uses an offline license key.

Sources

Facts on this page come from DeepSeek’s official API documentation and status page, plus a news report that is labelled as such, last checked November 19, 2026. These limits can change without notice, so we check the current notice and sources before relying on a detail.

Frequently asked questions

Is this the same as the 7-day or 72-hour long-chat notice?

No. The long-chat notice blocks continuing an older long conversation. This notice limits how much of an uploaded document DeepSeek’s website or app will read. The two started being reported around the same time, but they are different problems and have different next steps.

Has DeepSeek officially announced this limit?

Not as a help article or status incident that we can find. As of November 19, 2026, DeepSeek’s status page shows no incident announcing a 20-page or 10,000-token document cap. The wording quoted on this page is the notice users have searched for, and Caixin Global reported it on November 18, 2026, as a warning users had seen on the website and app since November 15.

Does the DeepSeek API have the same 20-page limit?

The published File Parsing documentation, checked on November 19, 2026, allows parsed documents of up to 1,000 pages and says the API does not silently truncate a document when the context is too long; the request fails instead. That is not the same as the website and app notice. API use is billed, and DeepSeek does not offer a consumer subscription for the chat app (Caixin, November 18, 2026).

If I split a 60-page file into 20-page parts, will DeepSeek read all of it?

DeepSeek has not published how many files the website or app will read at once, so we do not treat a stack of 20-page splits as a way to reconstruct the whole document in one chat. We extract the sections the next step needs, check that the notice does not appear, and add further sections only when we need them.

Can AI Chat Handoff read a DeepSeek document or chat?

No. Handoff reads ChatGPT and Claude conversations only, and it does not parse uploaded files. For this notice we extract the pages we need ourselves; once the work is in ChatGPT or Claude, Handoff can help move that conversation to a fresh chat later.

Continue with context

AI Chat Handoff is free in the Chrome Web Store.

Explore AI Chat HandoffAdd to Chrome — free

Related guides