Foundation models are AI models trained on enormous amounts of data that can subsequently be adapted to perform a wide range of tasks—much like an operating system that supports a variety of programs. One type is the language-aligned vision foundation model (VLM), which is trained on millions of image-text pairs, enabling it to connect images with language. Today, they are used in a wide range of applications, from image recognition and image search to image generation, and in combination with large language models.