When you submit a query with Phind, it will tell you which model it's using right underneath the query. At least it used to occasionally say that it's using GPT-4 by annotating it with (IIRC) "best model", but now I can only get it to reply with what it calls "Phind model". I don't know what that means or how it relates to GPT-3/4. Presumably they haven't created their own competitive language model from scratch without telling anyone, but who knows.

Though even when it was easier to get it to use GPT-4 it wasn't consistent, and would only use it when it inferred that the query was complex/technical enough.