Apple's multimodal large language model Ferret officially released as open source

October of this year.pomegranateand a team of researchers from Columbia University have released a multimodal large language model called Ferret. Compared to traditional models, Ferret is unique in its ability to accurately recognize and describe the content of an image, as well as accurately identify and localize various elements in the image.

The Ferret model is available in two versions, 7B and 13B, providing users with different options. In order to further improve the performance of the model, Apple has also built a large dataset called GRIT. The dataset contains 1.1M samples covering rich and diverse hierarchical spatial knowledge, providing strong support for model training.

At the time of its release, Apple provided only the code and weights, which were primarily geared toward research use rather than commercial applications. As a result, the news did not attract widespread attention. However, with a series of influential papers published by Apple a few days ago, revealing that theiPhoneThe major breakthroughs made in deploying large language models on a growing number ofAIExperts and researchers are beginning to take notice of the previously released Ferret model.

This series of papers not only demonstrates Apple's in-depth research and leadership in the field of AI, but also brings new thinking and inspiration to the entire industry. Through close cooperation with academia, Apple has successfully transformed its research results into actual products and services, bringing users a smarter and more convenient experience.

In the future, with the continuous development and improvement of multimodal large language modeling technology, we are expected to see more excellent models like Ferret emerge. These models will further promoteartificial intelligence (AI)The field of development, and in various fields to play its strong application value.