4o also added image input (previously only previewed in GPT4-vision) and enabled advanced voice mode audio input and output.