ChatGPT max PDF size is 512 MB, so why does a 4 MB file fail?

Size is rarely the real limit

The number is generous. A 500 page text-based PDF is typically a few megabytes, which is three orders of magnitude below the ceiling.

To actually reach 512 MB a PDF has to be carrying enormous images: a scanned book at high resolution, a photo book, a set of technical drawings. And that class of document has a different problem, because images are discarded before any reading happens.

So a PDF that genuinely hits the size limit is usually one ChatGPT could not have read anyway.

What makes a PDF large

The pattern is worth noticing. The things that make a PDF big are the things ChatGPT throws away. Text weighs almost nothing.

Shrinking one, and why it is free

Since images are discarded on every plan except Enterprise, compressing costs you nothing that was going to be used.

Compress the PDF. Nearly every PDF tool has a reduce-file-size option. It lowers image quality, which is the part being discarded.

Export to plain text. The most effective by a distance. A 60 MB PDF often becomes a few dozen kilobytes as text, and text is precisely what ChatGPT extracts. Keep the original for yourself if you need the layout.

Extract the pages you need. A chapter is smaller than a book and usually produces a better answer, because there is less irrelevant material competing for attention.

The two limits that stop people who are nowhere near 512 MB

Length. Text and document files are capped at 2 million tokens. A long PDF can be a couple of megabytes and be well past this. It uploads, ChatGPT reads what fitted, and nothing warns you. Check by asking what the final section is called.

Scanning. A scan has no digital text, so extraction returns nothing. Every plan except Enterprise behaves this way, which is why paying more does not fix it.

Between them, those two account for nearly every failed PDF. Size accounts for very few.

The other PDF ceiling

Separate from file size, there is a limit of about 100 images per conversation, and PDF pages count toward it when a PDF is handled visually.

That is where the number people have seen comes from. It applies per conversation, so starting a new chat resets it, and converting the PDF to text removes it from the equation entirely, since text is not images.