I have been ai studio for months now, the old sessions can attach very larger tokens as files but new sessions are complaining even at 12 k. Most strangely I managed to attach 12k and then tried to attach just over 11k and it is complaining that I should reduce the token size. Is there a new role out or something? As a matter of face my previous session seems dead in the water with token counts, but I can use much older session.
Hi @Kevin_Ehsani,
Welcome to the Forum,
Thank you for reporting. I was unable to reproduce it on my end when attempting to attach a PDF file exceeding 100K tokens. To assist with our investigation, please provide the file type, the precise size of the failed file and steps to reproduce the issue.
Hi @Kevin_Ehsani,
Thanks for the screenshot! I see the error. You’re using the Nano Banana model, which has a 32,768 token usage limit. Your current prompt, including voice_manager.py is 50,568 tokens, exceeding the limit and causing the Token count for the prompt has been exceeded. Reduce your prompt size to run message.
If a similar prompt worked before, it was either shorter or used a different model with a higher token usage limit.
To fix this, either shorten your current prompt to under 32,768 tokens or switch to a model with a larger token usage.
You are so right. I guess a new session uses nano banana not gemini 2.5 as in the past, something to pay attention to, I just took it for granted it is business as usual. Thanks a lot, goofy error on my part.
Hi @Kevin_Ehsani,
When you start a new session, you are indeed getting Nano Banana by default currently. If you want a different model (like Gemini 2.5 Pro, for example), you would then use the model selector on the right-hand side to change it.
Thank you!
@chunduriv I sent an example of a broken chat in another thread, the token loading bug in old, large chats is sill there. Please continue to look into this or at least give me an update on it.