Gemini 2.5 flash lite empty response after tool call

Hi,

I’ve recently run into a very odd problem when using the gemini 2.5 flash lite model.

After detecting the tool call and the necessary parameters for it, I send it back the response
and in return it generates an empty string.
This happens only with some tools more than others, and it is consistently reproducible
for specific inputs.

I’ve tried both the OpenAI SDK and the Vertex AI one and the result is the same.

Notably, none of this happens for the gemini 2.5 flash model, but I need the low latency
of the lite variant, so I’m trying to get to the bottom of this.

Any ideas on why this could happen or what we could try investigating?

Thank you,
Danut

Hi @mateidanut , Welcome to the AI Forum. Thank you for reaching out! To help us investigate why the model is returning an empty string after a tool call, could you please provide a minimal reproducible code example? This will allow us to replicate the issue and investigate further.

We’ve managed to find a solution in the end. The cause was the presence of unicode characters in the conversation, which we’ve since removed.