Gemini 2.5 flash native audio dialog (LIVE/Multimodal API) Tool Calling

Seems to be an ongoing issue, i see a few others posts about it, But No errors or anything popup the Model just refuses to acknowledge that it has functions available most the time, OR When it does, it will say “Okay! I’ll do this for you” after i say to do a function available, It will pretend it does the function but never executes it, It RARELY Works, I’d say about a 1 in 50 chance it actually works properly

2 Likes

Hi @BarricadeBandit,

Thanks for reporting. To help you better, please provide more information about your setup, specifically your use case and the number of functions you’re using.

Thank you!

Hello @chunduriv,

My use case was a Real Time conversational Chat bot, and the number of functions available to the LLM Is 10, Like i said, It rarely will actually call the Tools 99% of the time it will just continue the conversation as normal and act like it doesn’t know what the function is (as if it doesn’t even have functions available) The Older 2.0 Flash Live model works Perfectly 100% of the time with Function Calling, Same functions and all, The functions were pretty basic too, such as i have a function to get the Current time, Current Date, etc.

I can confirm that this happens to us as well, the function is called 1 out of 20 times for us, sometimes the model output audio of the function’s description instead. You can find a github issue of other people having the same problem here.