LLMs are renowned for using so much RAM. Does this mean we can't essentially use FoundationModels in extensions such as MessageFilterExtension? I assume the system kills the extension before we even get a response.
Thanks for this great question. The system language model (SystemLanguageModel) is not loaded into the app / extension's memory, and so using it doesn't count on the memory limit of your extension. If you are using your own on-device model, the model will be loaded to the memory of your app / extension, and so you will need to test if that is fine for your extension. Note that some extensions don't allow XPC due to privacy reason, and hence can't use a model via the Foundation Models framework.
Best,
——
Ziqiao Chen
Worldwide Developer Relations.