DiffThinker turns hard picture-based puzzles into an image-to-image drawing task instead of a long texting task.
COOPER is a single AI model that both “looks better” (perceives depth and object boundaries) and “thinks smarter” (reasons step by step) to answer spatial questions about images.