Visual inputs: The key feature of the newly released GPT-4 Vision is that it can now accept visual content such as photographs, screenshots, and documents and perform a variety of tasks. Object detection and analysis: The model can identify and provide information about objects within images.