GPT-6 has multimodal input, not output. It can call GPT images by itself so it's seamless, but it still isn't generating the images.