Guide

How Many PDF Pages Can Claude Read?

Last verified: August 2026

Short answer: Through Anthropic's API, Claude accepts a PDF of up to 600 pages per request — or 100 pages if the request's context window is under 1 million tokens — with a 32 MB total request-size cap. In the Claude.ai chat app, you can practically upload PDFs of around 100 pages / ~30 MB per file, and up to 20 files per conversation.

But page count is rarely the wall you actually hit. The context window usually is. Here's what that means, and what to do when your document is too big.

The limits, by where you're using Claude

If you're a developer using the API: the documented ceiling is 32 MB per request and up to 600 pages — but that 600 only applies when your request's context window is 1 million tokens or larger. On a smaller context window, the cap drops to 100 pages. Encrypted or password-protected PDFs aren't supported at all.

If you're using Claude.ai (the chat interface): in day-to-day use you'll find PDFs work best under roughly 100 pages and ~30 MB per file, with a limit of about 20 files per conversation. Go past the file-size or page ceiling and the upload is rejected before Claude ever sees it.

These numbers come from Anthropic's own PDF documentation and can change as models evolve, so treat them as current-as-of-mid-2026 rather than permanent. When in doubt, the API limits are the officially documented ones.

Why page count is the wrong question

Here's the part most guides miss: you can be well under the page limit and still get stuck.

When you hand Claude a PDF, every page is converted into an image and has its text extracted — both are sent to the model together. Each page typically costs somewhere between 1,500 and 3,000 tokens just for the text, plus additional tokens for the page image. A dense document — small fonts, packed tables, lots of graphics — can fill the entire context window long before it reaches the page cap.

So the practical limit isn't "100 pages" or "600 pages." It's how much of Claude's context window your specific PDF eats up. A clean, text-light 200-page report might fit fine. A 40-page slide deck full of dense charts might choke. This is also why Claude can fail on a large PDF even when the page count looks safe.

What actually happens when you go over

Two different failure modes, depending on which ceiling you hit:

File too big / too many pages: you get a rejection at upload — the file simply won't process.
Context window overflow: the file uploads, but the combined tokens exceed what Claude can hold at once, so the request fails or the document can't be fully read.

The fix for both is the same idea: give Claude less, but the right less.

How to get a large PDF into Claude anyway

You don't need the whole document. You need the pages that matter, in a form that fits. Three approaches, in order of how often they help:

1. Send only the pages you actually need. Most of the time you're asking Claude about a handful of pages inside a big file — a specific section of a contract, three slides, the results in a paper. Pull just those pages out and hand Claude a clean, focused packet. That's exactly what PDF Source Prep does: pick the pages you want across one or more PDFs, bundle them with a citation-ready prompt, and export a tidy source packet — all in your browser.

2. Split a long document into parts. If you genuinely need all of it, break it into chunks Claude can take one at a time. For unusually tall single-page PDFs (long web captures, chat exports, receipts), PDF Unroll slices them into clean, page-sized sheets.

3. Shrink the file. If you're bumping the 30 MB size cap rather than the page cap, the weight is usually embedded images. Downsampling them can bring a rejected file back under the limit.

Prep the pages before you upload

PDF Source Prep lets you pick just the pages you need from one or more PDFs and export a clean, citation-ready packet — sized to fit, not just trimmed at random. It runs entirely in your browser; nothing is uploaded.

Try PDF Source Prep — free No sign-up · 100% in your browser

A note on confidential documents

If your PDF is a client contract, a financial statement, or anything under compliance rules, how you split it matters as much as whether you can. Every tool on this site runs 100% in your browser — nothing is uploaded to a server, nothing is installed. You trim and prep the file locally, then decide what to hand to Claude. For legal, finance, and healthcare work, that's not a nice-to-have; it's the whole point.

Frequently asked questions

Can Claude read a 500-page PDF?

Via the API on a 1M-token context window, yes — up to 600 pages, size permitting. In the chat app, a 500-page file is over the practical ceiling; split it or extract the sections you need.

Does Claude read the whole PDF or just the first few pages?

The whole thing, as long as it fits within the size, page, and context limits. Claude doesn't silently truncate to the first pages — if the document is too big, it fails rather than partially reads.

Why does Claude say my PDF is too large even though it's under 100 pages?

Almost always the context window or the 30 MB size cap, not the page count. Dense pages and embedded images cost far more tokens than plain text, so a short-but-heavy PDF can overflow before a long-but-light one does.

Is it better to paste the text or upload the PDF?

If you only care about the words (no charts, tables, or layout), pasting text sidesteps the file-size limit entirely and is often lighter on the context window. Upload the PDF when the visual layout matters.

Limits described here are based on Anthropic's official PDF documentation and were last verified in August 2026. Anthropic may change these caps as new models ship — check the official PDF support docs for the latest.

PDF Source Prep · a free, browser-only tool for prepping PDFs for Claude and ChatGPT.
Your files never leave your browser.