GPT-4o Responses Contains Combined Unreadable Characters

I have a web that is using assistants for having a conversation. When I changed to use GPT-4o model, now I am usually receiving non-standard characters or combined languages different than english. For example:

This is my assistant configuration:

Try a lower temperature to begin

1 Like

Sure, will try that! I thought adding +0.25 was not enough to having this type of responses.

AFAIK it shouldn’t be, but this output seems very similar to high-temperature insanity

1 Like