Phi-3-vision is the first multimodal model in the Phi-3 family. It combines text and image capabilities, allowing it to reason about real-world images and extract and understand text from images. The model has been optimized specifically for understanding charts and diagrams. It can generate insights and answer questions related to charts and diagrams. Phi-3-vision builds on the language model Phi-3-mini but adds image understanding capabilities while still being a relatively small model size.