Difference in results between Gemini via interface and package

[louisarthur]

Hello everyone, I'm facing a problem in relation to using Gemini through the interface and the package, using default parameters to analyze an image and using the same prompt, I'm having a difference in the response. In the image detection package it is X and in Gemini through the interface it is Y. The interface always gets it right, the package gets it wrong.
Anyone else facing the same problem?

[hortigado]

I am facing the same problem, the responses through the API are poor, sometimes it throws errors and many times you get an empty response.

[natebosch]

Gemini through the interface

Do you mean gemini.google.com or the mobile app? The Chat interface to gemini?
I don't think that is guaranteed to be backed by any specific model that you can access through the rest API.

the responses through the API are poor, sometimes it throws errors and many times you get an empty response.

These all look like questions that are best directed to the folks who maintain the model and the backend.

@singhniraj08 - this seems to be a question about the backend behavior. How do we typically route those issues? I don't have knowledge about the model behavior or the alternative interfaces to the LLM.

I don't see anything in this issue which is actionable in this repo.