Ferret-UI is a novel MLLLM designed to enhance the understanding of mobile UI screens. It possesses the capabilities of referencing, locating, and reasoning, enabling it to handle tasks on UI screens. Utilizing its "Any Resolution" technology, it effectively addresses the issue of small object recognition by magnifying details. Its performance is particularly notable in advanced tasks involving conversational abilities, showcasing its strengths in task formulation, training sample collection, model architecture establishment, and dataset creation.