Apple reportedly develops local AI model running applications

According to Gelonghui, sources have revealed that Apple is developing a Mac application named Harbor, which can turn computers equipped with Apple chips into private AI inference servers, supporting direct downloading, management, and running of open weight models without the need to upload data to the cloud.

The application can also connect two or more compatible Macs via Thunderbolt 5, automatically forming a cluster and integrating unified memory to run large models that exceed the capacity of a single machine or handle more concurrent requests without manual configuration or using the command line. The application is currently in a non-public preview by Claris, a software company under Apple, and the official launch date has not yet been announced.