Apple has joined the AI competition by releasing OpenELM, its new open-source language model that can run directly on devices instead of using cloud services. OpenELM, short for Open-source Efficient Language Models, is created to work on devices instead of the cloud. You can find these models on the Hugging Face Hub, a popular platform for sharing AI code.
Apple’s LLM includes eight language models, four pre-trained with the CoreNet library and four fine-tuned models. They’re using a layer-wise scaling strategy to improve both accuracy and efficiency, as per the release white paper PDF.
To differentiate its LLM from competitors, Apple has not only released pre-trained models but also the entire framework, which includes code, training logs, and multiple versions.
Apple chose to make the OpenELM models open source to empower and enrich the research community with state-of-the-art language models. This decision allows researchers not only to use the models but also to study their internal workings. This leads to quicker progress and “more trustworthy results” in natural language AI, as stated by Apple.
Researchers, developers, and companies can use Apple’s OpenELM models as they are or customize them to fit specific needs. This is different from before when companies usually only shared the model weights and code for using the model, without giving access to the training data or settings used to train the model.
OpenELM protects your privacy and reduces the risk of cloud server breaches by keeping data and processing on your device. It also works offline and is more accurate and efficient than similar models, according to Apple.
Open sourcing benefits both researchers and Apple. It allows Apple to collaborate with the research community, letting others improve OpenELM. This openness also helps attract top talent to Apple. Apple sees OpenELM as a way to drive AI advancements for everyone, not just the company.
Although Apple has not yet introduced these AI capabilities to its devices, the release of iOS 18 is imminent, and rumors suggest Apple plans to bring on-device AI features with the new OS. With the launch of its own LLM, it’s clear that Apple is preparing to upgrade its devices, including iPhones, iPads, and Macs, with AI.
Apple may put its large language models into its devices, making user experiences more personal and efficient. This could also help keep user information private and give developers better AI tools.
