ChatGPT PDF limits by plan: file size, token caps, Visual Retrieval, and Projects
Updated: 3 days ago
A PDF can be small in megabytes and still be difficult for ChatGPT because file size, token count, project quotas, storage quotas, and visual processing are separate limits. The current hard ceiling for a file uploaded to ChatGPT is 512 MB, while text and document files are also capped at 2 million tokens per file.
THE CURRENT HARD LIMITS
Each uploaded file has a 512 MB hard limit. Text and document files, including PDFs, also have a 2 million-token cap per file. These are different constraints: a PDF can fit under 512 MB and still exceed the text-token ceiling.
··········
PROJECT LIMITS DEPEND ON PLAN
Project file counts are currently capped at 5 files for Free, 25 for Go and Plus, and 40 for Edu, Pro, Business, and Enterprise. These limits describe files attached to one Project, not the maximum number of pages inside a PDF.
........
Constraint | Current rule | What it affects |
Per-file size | 512 MB | Whether the file can upload |
Text/document size | 2M tokens | How much document text one file may contain |
Projects: Free | 5 files | Files attached per Project |
Projects: Go/Plus | 25 files | Files attached per Project |
Projects: Edu/Pro/Business/Enterprise | 40 files | Files attached per Project |
........
··········
PDF VISUALS ARE PLAN-DEPENDENT
The largest functional difference is not the megabyte ceiling but visual retrieval. Enterprise can analyze images, diagrams, and charts embedded in PDFs uploaded in prompts. Other plans use text-only retrieval for document files, so embedded visuals can be ignored even when the PDF itself uploads successfully.
··········
UPLOAD RATE AND STORAGE ARE SEPARATE
File-upload rate limits and storage quotas can stop an upload even when the PDF itself is valid. Current guidance allows up to 80 file uploads every three hours, while Free users are limited to three file uploads per day. Shared storage caps also apply across uploaded files.
··········
WHY PAGE COUNT IS NOT A RELIABLE LIMIT
PDF pages vary enormously in information density. A 300-page slide-like document can contain less extracted text than a 100-page legal filing. Token count and layout complexity therefore describe practical difficulty better than page count alone.
........
Problem | Likely constraint | First action |
Upload rejected immediately | File size/type | Check the file itself |
Few uploads but limit error | Rolling quota/storage | Check quota and storage |
Charts ignored | Text-only retrieval | Upload visuals separately or use supported Enterprise flow |
Long PDF answers seem incomplete | Retrieval/context | Ask section-specific questions |
........
··········
DATA STUDIOS LIMIT STACK
For troubleshooting, test limits in this order: file size → document token count → upload-rate quota → Project file count → account storage → plan-specific visual processing. This avoids treating every failed PDF upload as the same problem.
··········
WHAT THE LIMITS MEAN IN PRACTICE
For ordinary text-heavy PDFs, the 2 million-token document ceiling is usually more informative than megabytes. For scanned or visual PDFs, the critical constraint is often not the nominal file limit but whether the plan can actually interpret the information embedded as images.
··········
FOLLOW US FOR MORE
··········
··········
DATA STUDIOS
··········