GPT-4 is a large multimodal model (accepting text or image inputs and outputting text) that can solve difficult problems with greater accuracy than any of OpenAI's previous models, thanks to its broader general knowledge and advanced reasoning capabilities.