That’s because OpenAI trained a new model end-to-end that works across text, vision, and audio. Previously, GPT-4 would use separate models for audio, text, and images. With everything ...