In October 2023, Apple, jointly with researchers from Columbia University, discreetly released an open-source multimodal Large Language Model (LLM) named 'Ferret'. While it came with code and weights, Ferret was released for research purposes only on 30 October without any commercial license. Ferret is capable of 'referring and grounding anything anywhere at any granularity' in an image. It may also accomplish this by utilising any form of area inside a picture.

Why does it matter?

Ferret's release is seen as a significant move for Apple, which has traditionally been known for its secrecy and closed proprietary systems. The news did not initially receive much attention, but the situation changed recently due to increased interest in open-source models and the potential for local LLMs to power small devices. The interest was further fueled by Apple's announcement of a breakthrough in deploying complex AI systems, such as LLMs, to run on iPhones and iPads.

While Ferret is open-source, it is important to note that it is released under a non-commercial license. This means it cannot be commercialised in its current form. Despite this, there is potential for it to be used in a future Apple product or service.