Librechat started using the most expensive gpt 4.5 mid conversation and spent all my $ | LibreChat | Page 1

lime condor Mar 11, 2025, 9:25 PM

#

I'm using openrouter. So I was using my default profile deepseek-r1, then switched to flash2.0 mid conversation.
Please take a look at the log. You clearly can see it is the same conv id. And I don't have gpt4.5 anywhere in my profiles. I never set it anywhere as default.
That is super frustrating. Happened second time and like 15 bucks is gone.

📎 message.txt

minor sandal Mar 11, 2025, 9:26 PM

#

@pastel violet !!!!!!!^

#

@noble marsh ^^^

pastel violet Mar 12, 2025, 12:12 PM

#

lime condor I'm using openrouter. So I was using my default profile deepseek-r1, then switch...

did you attach images to this conversation?

#

likely you did, since gpt-4.5 is a vision model, it was probably used as a fallback model for the images

pastel violet Mar 12, 2025, 12:27 PM

#

lime condor I'm using openrouter. So I was using my default profile deepseek-r1, then switch...

anyway, sorry this happened to you, I will change the order vision models are used for the cheapest options to be considered first. If you'd like an API key with $15 credit for openrouter, send me a DM

lime condor Mar 12, 2025, 7:13 PM

#

@pastel violet yep, I did. That was actually the first thing I suspected, but I wasn't sure. Anyway, I'm glad we've figured that.

#

actually it might be a good idea to have a setting to set a fallback model.

#Librechat started using the most expensive gpt 4.5 mid conversation and spent all my $