Gemini Flash 1.5-002 suddenly shorter context window (1M -> 32K)

[Sangjae]

Hello,

I got error when I called VertexAI API with gemini-1.5-flash-002.

Error message is "400 The input token count (32920) exceeds the maximum number of tokens allowed (32768)."

Size of context window of gemini-1.5-flash is 1M, but this error message shows context windows is decreased to 32K.

(Update 24th Mar - This error is still occurring)

[caryna · Accepted solution]

Hi @Sangjae,

Welcome to Google Cloud Community!

I recommend submitting an issue report regarding this so that our Engineering Team can look into it. Before filing, please take note on what to expect when opening an issue.

For further clarification, you may reach out to Google Cloud Support for a one-on-one discussion about your issue. Their team has the specialized expertise to diagnose and resolve complex problems effectively.

Was this helpful? If so, please accept this answer as “Solution”. If you need additional assistance, reply here within 2 business days and I’ll be happy to help.

[Sangjae]

Thank you for kindly reply!

I'll submit an issue report.

[espenb]

Experiencing the same issue randomly, with flash 2.0. It works fine for most requests, but suddenly we get an exception with the same message you're referring to – a token limit of 32K. 

Did you find a solution, @Sangjae?

[Sangjae]

I just heard from Googler that problem was just solved. And I don't get same issue now.