Formats and limits
Input: what you can send
| Type | Extensions | Notes |
|---|---|---|
.pdf |
Digital PDFs are read directly; scanned PDFs go through OCR automatically | |
| Word | .docx |
The binary .doc format is not supported (see below) |
| Rich text | .rtf |
Formatting is stripped, text is kept |
| Plain text | .txt, .csv, .log |
Read as text |
| Images | PNG, JPG, JPEG, WEBP, TIFF (.tiff, .tif), BMP |
Always read with OCR |
A file with an unknown extension is read as plain text when possible.
Output: what you get back
format |
data contains |
|---|---|
json |
A JSON object. With a schema, an object that follows it |
markdown |
A string of Markdown |
html |
A string of HTML |
The format parameter is not case sensitive. Any other value is refused with 400.
Limits
| Limit | Value |
|---|---|
| File size | 50 MB per file (413 if exceeded) |
| Batch | Up to 100 files and 200 MB in total per request (413 if exceeded) |
| Text sent to the AI extractor | The first 25,000 characters of a document; when the text is longer the answer has input_truncated: true |
| Conversion request rate | 20 requests per minute on the free tier; higher tiers allow more (429 with Retry-After when exceeded) |
| Monthly quota | Counted in pages, per calendar month (UTC). The free beta includes 50 pages per month and 500 welcome pages that do not expire; 402 when used up |
| Background job results | Kept for 24 hours |
How each limit behaves, and how to stay within it, is explained in Rate limits, Usage and quota and Batch and ZIP.