Apple’s new AI model hints at how AI could come to the iPhone

It has four sizes: 270 million parameters; 450 million parameters; 1.1 billion parameters; and 3 billion parameters. Parameters refer to how many variables a model understands in decision-making from its training datasets. For example, Microsoft’s recently released Phi-3 model bottoms out at 3.8 billion parameters, while Google’s Gemma offers a 2 billion parameter version. Small models are cheaper to run and optimized to work on devices like phones and laptops.

Apple CEO Tim Cook teased that generative AI features will be coming to the company’s devices, saying in February that Apple is spending “a tremendous amount of time and effort” in the space. However, Apple has yet to share specifics on what its use of AI might look like.

The company has released other AI models before, though it hasn’t released any AI foundation model for commercial use like its competitors have.

In December, Apple launched MLX, a machine learning framework that ideally makes it easier for AI models to run better on Apple Silicon. It also released an image editing model called MGIE, which lets people fix photos with prompts. Another model, Ferret-UI, could be used to navigate smartphones. Apple is also rumored to be working on a code completion tool similar to GitHub’s Copilot.

However, even with all the model releases from Apple, the company reportedly reached out to Google and OpenAI to bring their models to Apple products.