A multimodal workhorse that combines vision and language understanding in a single model. It can look at an image and reason about its contents in natural language, making it capable of tasks like visual question answering and image description. Being open-weight means you can run it locally, though the Llama 2 license does impose some usage restrictions.