Apple has kept its generative AI strategy under wraps, but recent developments indicate a shift towards deploying AI technology directly on its devices. The company introduced OpenELM, a collection of four streamlined language models available on the Hugging Face platform, tailored for efficient text processing such as email composition. OpenELM stands for "Open-source Efficient Language Models" and is freely available to developers.
These models come in various capacities—270 million, 450 million, 1.1 billion, and 3 billion parameters—allowing them to handle different levels of complexity in data processing. Parameters in this context measure the breadth of data a model can analyze to make decisions. Smaller models, which are less costly to operate, are specifically designed for efficiency on personal devices like smartphones and laptops. Previously, Apple has released several AI models, though none as foundational for commercial use as those from some competitors.
Last December, Apple also launched MLX, a framework to enhance AI performance on Apple Silicon, alongside MGIE, an image-editing tool that edits photos via user prompts, and Ferret-UI, which is designed to improve smartphone navigation.