Try out "gpt-5-chat" model. I can't tell if it is the same but it is running a hell of a lot faster.