Hello,
I’ve read about new PDF support in the Gemini API however I can’t find any documentation for it.
Has it been released yet?
1 Like
Do you have a reference to where you heard about PDF support?
Yes, Logan mentioned it on his X:
To anyone this may help:
I found the new documentation (I think), its subtle but it’s there and I’m pretty sure it didn’t exist before:
For my personal project, I ended up going another route due to the nature of the project.
I’m using context caching and need really low latency in response times.
So, for me, it was better to convert my PDFs to text using a tool like PyPDF2 and then only have important figures from the PDF as images included in the system prompt.
PDFs in Gemini are just converted to images anyway, so if you need low latency I’ve found that keeping the token count under ~600k is best.
I’ve noticed high latency when the token count is higher than 600k.
And there are frequent response errors when the token count is higher than 2M.
Hopefully this helps someone 