Compare image-text models like CLIP to image-only models and explain when you would use each at OpenAI.
