I understand Google is having an issue and gemini-2.5-pro isn’t available.
I never asked for requests to be routed to gpt-4.1 in this case.
I will pick the model I want to route to when there is a failure.
This is a poor decision, please make this configurable or disable it.
10 Likes
request id: 0f90fdfb-32b4-4ff7-851a-d6dc10890f89
I am happy to see the error message got updated. An automatic fallback is not an option, though. different models behave very differently. not only might this lead to messed up results, it also costs money for an unrequested service.
1 Like
I disagree, so I suggest a compromise:
Add an option: when the model call fails, automatically switch to a backup model.
1 Like
no, like I said: this will lead to a waste of time as the user will not only probably pay something unrequested – he will also need to invest time and money again to create the initially desired state of the codebase.
Instead:
prompt the user, let him select from a list of fallback models, so the user can decide if he wants to take the risk of using a fallback model or not.