By skyforbes 
I have been loving Gemini 3 since release. But I've tried to get it to implement gpt5nano in a project and it just simply refuses. I even gave it an API Key to use.
A bit strange, actually, but it insists on Gemini models. I config the settings myself in a prior turn, but after working on it a bit I noticed it mentioned Gemini so dug deeper and saw it had replaced gpt5nano implementation with 3 pro and 2.5 flash. Now I worry it will try and "fix" it again and replace with Gemini if I revert.
Anyone else see this weirdness? Also, I asked it a question, but it lost the thread for a while and tried to do a bunch of other work, before coming back with an incorrect and misguided response.