I do feel that the 4o image generation demonstrates something really different happening that's eye opening. It can translate objects from source images into new images, it can maintain style, it can write proper text. By increasing the reasoning it actually *understands* what's happening in the img