I’ve been using the Live Multimodal API for a project where an LLM controls a robot. The setup relies on function calls being triggered in certain situations. As of now, these function calls have completely stopped working, not just in the current version, but across multiple previous versions of the project that were working fine before. This strongly suggests the issue is on the backend.
Is anyone else experiencing the same problem?
You’re far from the only one : https://discuss.ai.google.dev/t/multimodal-live-models-failing-tool-call/83017/21
Hi @Umur_Yildiz,
Welcome to the Forum,
To better assist you, please provide more information about your setup. This includes details such as your use case, model name and the number of functions you have defined.
Thank you!
Hi @Umur_Yildiz,
The gemini-2.5-flash-native-audio-preview-09-2025 model offers significantly improved function calling. Please reattempt your task with this new model and let us know your thoughts.
Thank you!