How many files can you put in a ChatGPT Project? 5, 25 or 40

The numbers

Every plan is also capped at 10 files uploaded simultaneously inside a project, which is a queueing limit on one batch rather than a ceiling on the project, and is frequently mistaken for one.

This is one of the few places where the plan you are on genuinely changes an upload limit. Most of the others, the size ceiling and the length cap, are identical everywhere.

Three different limits are in play

Worth separating, because hitting one feels much like hitting another.

The project count. 5 files per project, 25 files per project or 40 files per project, per project. Hit it and you cannot add another file to that project until you remove one. It does not come out of your chat allowance: a Free user gets 5 files per project on top of, not out of, the 3 uploads a day.

The rolling upload rate. 80 files every 3 hours. Uploading into a project is still uploading, so a bulk import of thirty files spends thirty slots.

Storage. 25 GB per user, shared across every project, every chat and every custom GPT. Cumulative, and it never clears with time. There is also a separate and much smaller Library quota, 500 MB on Free and 20 GB on Plus and Business, which is what Settings then Storage actually reports.

Making the allowance go further

The ceiling counts files, not content, so combining is almost always the answer.

Group related documents into one file. Ten well organised files carry far more usable knowledge than twenty arbitrary ones, and the project stays easier to reason about.

Put a clear header before each original document when concatenating, so the boundaries survive:

=== Supplier agreement, 2026 ===
...

=== Amendment 1, March 2026 ===
...

Convert to plain text. Smaller, cleaner, and it strips images that were going to be discarded anyway on every plan except Enterprise.

Take out what nobody will ask about. A tighter project produces better answers, because there is less irrelevant material competing for attention on every question.

Files in projects are visual too, and that matters

The same rule applies inside a project as anywhere else: except on Enterprise, ChatGPT extracts digital text from a document and discards the images.

So a project full of scanned reference material contributes nothing, and nothing will tell you. The project looks well stocked and answers vaguely.

If a document matters and its content is in images, OCR it before adding it, or the slot is wasted.

Projects against custom GPTs

Both hold reference files, and people mix up their limits.

The custom GPT limit is the stricter of the two in practice, because it does not refresh at all. It is also the one most readers will never meet: GPT creation and publishing have been withdrawn from all personal accounts, Free, Go, Plus and Pro, and survive only in Business, Enterprise and Edu workspaces.