GPT-4o, new flagship model which can reason across audio, vision, and text in real time: