ChatGPT File Upload Limit

You can use ChatGPT to compare work documents and analyze spreadsheets, but getting a batch through starts with identifying which limit applies. Documents can be up to 512 MB, text and documents also face a 2 million-token cap, and Free accounts get 3 file uploads per day. A smaller file will not fix an exhausted upload allowance.

Checked November 7, 2026. The numbers below come from OpenAI's Help Center. Where it does not publish a fixed number, the table says so.

ChatGPT File Upload Limits by Plan

Free has a daily upload cap; the general published allowance is up to 80 files every 3 hours. OpenAI may reduce those limits during peak hours. Its upload article does not give Business or Enterprise a separate, higher numerical allowance.

Sources by column: rate from Uploading files and audio to ChatGPT; conversation count from Data analysis with ChatGPT; Project capacity from Projects in ChatGPT; saved-file storage from Using Library to manage files.

The paid-plan entries use the upload article's general ceiling, not a guaranteed allowance for every account or model. It does not publish a separate paid-plan daily total. Do not multiply the rolling allowance into a promised daily quota.

Storage has another limit: the upload article lists shared upload storage of 25 GB per user and 100 GB per organization. These are distinct from the plan-specific Library figures above. The Help Center presents them separately; it does not explain every interaction between the allowances. Check the storage screen in your account rather than adding them together.

File Size: PDFs, Spreadsheets, and Images

The file's type determines which size test it must pass. These published limits are not larger just because your plan has more Project slots.

Source: the upload article linked above. Image-count behavior comes from the Image Inputs FAQ, which supports PNG, JPEG, and non-animated GIF files and advises reducing image quantity or size if problems occur.

A token is a small unit of text the model processes, not a page. Think of the size cap as the width of a loading gate and the token cap as the amount of written material inside the shipment. There is no single page-count conversion you can safely use for every PDF.

Conversation, Project, and GPT Limits Are Different

A Project's file allowance is not the number you can attach to any conversation. The data-analysis article says conversation limits vary with upload type, model, plan, workspace settings, and remaining allowance. It supplies no universal count.

Projects allow 10 files to be uploaded at the same time. That is a batch limit inside Projects, not a promise that Free users can upload ten files today. Their daily allowance still applies.

You can create unlimited Projects. Each holds up to 5 files on Free, 25 on Go or Plus, or 40 on Pro, Business, Enterprise, or Edu. The Projects article recommends removing unnecessary files, combining file data, or splitting work across Projects when you reach a Project's limit.

Custom GPT knowledge supports up to 20 files, with a maximum of 512 MB per file, subject to the other relevant restrictions. Knowledge means reference material attached to that GPT, not a quota for every chat you have with it.

Do not create a new GPT as your assumed escape route. The current Creating and editing GPTs guide says personal accounts cannot create or publish new GPTs. Editing existing GPTs depends on plan and permissions; managed workspaces must also follow their availability and transition notices.

For recurring document work, start with the ChatGPT Projects setup guide. Its organizational benefit is keeping the relevant files, conversations, and instructions together.

What to Do When You See “Upload Limit Reached”

Match the fix to the constraint before retrying. OpenAI says failed upload attempts can sometimes count toward the rate limit, so repeated attempts can make diagnosis harder.

1. Confirm your account and plan. Check that you are signed in to the account or workspace you intended to use.
2. Read the retry time. If the message gives one, wait until then. The upload article describes a rolling rate allowance but does not give a universal reset clock or promise that Free uploads reset at local midnight.
3. Inspect Settings > Storage on web. Storage usage is visible; your remaining rolling upload-rate quota is not.
4. Check where the saved copy lives. Deleting a chat does not delete a file saved separately in Library or Space. Delete that copy separately if you want to remove it. Remove GPT knowledge in the GPT builder.
5. Check service status, then escalate if needed. The upload article directs you to OpenAI Status. For support, collect the account email, screenshot, timestamp and time zone, browser or platform, and request ID if available.

Those are the upload article's troubleshooting steps. Storage cleanup addresses stored files; it is not a documented reset for the rolling upload allowance.

Split or Compress Only for the Problem It Addresses

Splitting helps when analysis is incomplete; it is not an upload-rate workaround. The data-analysis article recommends asking about specific sheets, rows, columns, or sections, or splitting a file into smaller files when the result is incomplete.

For images, the Image Inputs FAQ recommends reducing quantity or size. For a Project that is full, use the Projects article's options: remove unnecessary files, combine file data, or divide the work across Projects. Any resulting file still has to fit its applicable limits.

The cited pages do not describe general PDF compression as a way to bypass the token cap or upload quota. They also do not promise that a ZIP archive bypasses file-count limits. Avoid treating either as a guaranteed fix.

Use the API for a Separate Workflow

An API integration is a different file-processing workflow, not extra capacity inside your ChatGPT subscription. The API is the interface an application uses to send work to a model. OpenAI's upload article explicitly says its requirements, supported purposes, and limits are separate.

For a developer building recurring document intake, the current API file-input guide distinguishes direct file inputs, File Search for retrieving relevant material from large files, and Hosted Shell for detailed spreadsheet work. A developer can upload a file and reference its file ID, the handle assigned to that upload, in a model request.

Use the limits for the chosen API route. Do not carry ChatGPT's per-file allowance or Project count into that design. This guide keeps its numerical limits to the Help Center's ChatGPT figures.

Five Workflows Worth Preparing the Files For

Recurring spreadsheet and document work benefits most from getting intake right. These are proposed workflows, ranked by practical usefulness, not measured time savings.

The upload article documents these underlying tasks. The data-analysis article recommends structured spreadsheets and warns that scanned tables or complex layouts may not yield reliable exact values. When exact numbers matter, provide the spreadsheet or text data.

Directly attached PDFs can be analyzed for both text and images, including charts and diagrams. Files saved as Project sources or GPT knowledge are handled differently, and their embedded images may not be included automatically. Uploading successfully is not proof that every part was analyzed.

Two Small Tools a Team Could Build

The strongest opportunity is a file-readiness checker, because it can catch preventable errors before an upload. Search demand supports investigating the problem; it does not prove people will pay.

A local upload preflight checker. An operations team could use a browser tool that checks selected files against a maintained table of documented types and size caps. DataForSEO estimates 720 monthly searches for “chatgpt file upload limit” in the site's configured market. The smallest useful version would flag oversized files and explain whether the issue concerns file size, Project capacity, or an account allowance. An internal IT team is the likely buyer if upload support consumes staff time. The catch: a local checker cannot know a user's remaining ChatGPT quota, and OpenAI does not expose that remaining rolling allowance in the interface. It must never claim that an upload will succeed.

A PDF intake preparation tool. A research or procurement team could use a tool that selects relevant sections, preserves source filenames, and prepares smaller documents for focused review. DataForSEO estimates 720 monthly searches for “can chatgpt read pdf” in the same configured market. A minimal version would preview the selected sections and let the operator confirm what gets exported. The catch: splitting can remove context and creates more files to upload. The product would need to earn its place through reliable preparation, not a promise to bypass limits.

These are overlapping informational search audiences, not customer counts. Keep the first product local and narrow until a team demonstrates a recurring need.

Common Questions

How many files am I allowed to upload to ChatGPT?

Free users are limited to 3 uploads per day. The general published ceiling is up to 80 files every 3 hours, and OpenAI may lower it during peak hours. There is no single fixed conversation count in the current data-analysis article. Projects and GPT knowledge have separate file counts.

How big of a PDF can you upload to ChatGPT?

A PDF falls under the document limits: up to 512 MB and 2 million tokens per file. Passing those limits does not guarantee complete analysis of a complex or image-heavy document.

Why isn't ChatGPT allowing me to upload PDFs?

Check the account and plan, any retry time, the document's limits, and saved-file storage. Failed attempts can sometimes count toward the upload allowance. If those checks do not explain the problem, follow the upload article's service-status and support steps.

How can I bypass the file upload limit in ChatGPT?

The cited Help Center pages provide no universal bypass. Use the remedy for the limit: wait for an upload allowance, manage stored files, reorganize a full Project, or split material for incomplete analysis. None is documented as removing all the other limits.

Your Monday Move

Prepare the next batch around the question you need answered. For a weekly reporting job, use a structured spreadsheet with clear headers, decide which columns matter, and check its size before uploading. Put recurring reference material in a Project within your plan's allowance. After analysis, review the calculation and the source data.

That changes the workflow from uploading everything and diagnosing the failure afterward to preparing the relevant material, checking the constraint, and reviewing the answer. The benefit is fewer avoidable retries. No fixed number of hours saved is established by these sources.

If document intake keeps interrupting your team's work, build a repeatable file-processing workflow.