Visual examples from the Kosmos-1 paper show the model analyzing images and answering questions about them, reading text from an image, writing captions for images, and taking a visual IQ test ...