I am working on an agentic application and use OpenAi API for powering the application. I need to know the exact file upload limits (size, number, etc) for all kinds of files supported by OpenAI. For this I need an answer to these questions.
Are the 10-file upload limits only for the ChatGPT UI, and not enforced by the model itself when using the API?
When calling the model via the API, is there no hard limit on the number of files you can include in a single request (aside from context size and file size)?
What are the official backend/API limits for file uploads (file count, file size, total data)?
Here, I am talking about the chat completions API not the FIles API
Thank you very much for replying to the post, but my question still remains the same. Even if I use the Responses API it has no mention about limits of file uplaods (number of files, size, etc) in a single call
The only “upload” method on Chat Completions is an “input_file” user message attachment of a PDF file content part, which then has both text extraction and rendered page vision extraction performed on it.
You can upload multiple files, each less than 50 MB. The total content limit across all files in a single API request is 50 MB.
The actual tokens that will be provided are a bit opaque, but you can figure that a page will top out at about 1536 model tokens of context window consumption per image. There was documentation of a maximum 100 pages per file would be extracted, but that now seems elusive. Then you have the text component. This also cannot exceed the particular model’s input limit. You can “count tokens” on an API request with a special endpoint if you send in the Responses API shape.
Thanks for this one, can you please provide any official source that states this limit. Cause as far as I know this limit is stated only in case of the Responses API and not for Chat Completions API