Using "GPT:" is asking for "gpt-4o-mini". Unable to use with offline model.