A VLM (Vision Language Model) is a model that can look at an image and answer in words. "Describe what is in this photo." "Read the trend in this chart." "Where on this screen do I tap?"

It became possible when two currents met. On the language side, large language models (LLMs) had acquired enough reasoning ability. On the vision side, CLIP had shown how to turn an image into a vector that sits comfortably next to language. All that was left was the question of how to join the two.