OpenAI may have called GPT-5.2 its "most advanced frontier model for professional work," but tests conducted by the