I wonder what’s “my” file upload limit.
I uploaded some files in a chat, then in another chat i uploaded some more.
Why don’t we have a counter for all ChatGPT limits, so we can manage them better?
And what they mean about “later”? I tried 1 hour later and still this.
Thanks for sharing.
Haven’t seen this before either.
It makes sense that there are limits in place.
As far as we know these are also adjusted dynamically, at least for messages, based on overall usage from all ChatGPT users.
I agree, it would be nice to have more insights into the limits.
I tried now, 6 hours later and the chat let me upload again. I completed what I was doing, but I confess that I was worried about when it will limit me again, it’s quite unclear and doesn’t seem to fit into any of the documented limits:
File uploads FAQ | OpenAI Help Center
But it seems a non-documented limitation, see what happens:
I’m using chatgpt (model 4o) to summarize textbook chapters. Each page is in high quality jpg format. each chapter is 12 pages (12 jpg files of approximately 500 to 700kb each).
Chat has a limit of 10 files per upload. So for each chapter I send 10 and tell them to wait for instructions, then send the last 2 along with the prompt.
The first 2 chapters work very well, but from the third (when I already have 36 images sent in the session and 4 prompts), I start to see unspecified error messages without the generated response, so I tell it to regenerate and it starts to summarize previous chapters , or 2 chapters in the same answer, partially ignoring the instructions or getting confused.
At this point I saw a better result to get out of the problem by continuing the conversation, and asking him to correct the error, than by starting a new session which usually generates summaries in a different style and format than the previous chapters.
…
Since I faced errors and confusions, could that new “file limitation” because of excessive (how much?) use of image OCR (Vision) and processing? Maybe could not be the case if I upload text/doc files?
I’d expect that the whole conversation is fed to the model each message send and as you keep sending more chapters confusion ensues.
This would also be the case when starting a new conversation but the model finds that there are more images than those you are referring to in your message.
I don’t recall anybody else reporting a similar issue. If you have that many books to process maybe it’s worth not paying for plus and instead using the vision API?
Ps. As of today I don’t even know if one has to pay for plus to use this feature. Probably not? 
Ps 2. You can try this: instead of sending a new message with new images, edit the previous message and delete the old photos before adding new ones.
I forgot to mention this: I tried, but they don’t let me edit prompts when i uploaded files. That little pencil doesn’t appear for prompt ballons containing image thumbs together (above). I cannot edit or delete prompt, neither uploaded images, only entire chat session. Prompts without files I can edit as usual.
May I have to develop an interface to use Vision API, or there’s some official web ui for that?
Edit: checked Playground just now, it shows me only 3.5 models. Does that show other models if I add a payment method?
Alright. I was able to edit messages with images in the Android app but that’s likely not the best solution then.
You can try the playground:
https://platform.openai.com/playground/chat?models=gpt-4o
But you will need to pay a bit in advance.
Here in the forum you are encouraged to build your own interface:
https://platform.openai.com/docs/guides/vision
But maybe the playground will suffice.
This topic was automatically closed 2 days after the last reply. New replies are no longer allowed.