This post is from the WWDC26 Metal Q&A.
Are there opportunities for developers to use Apple Intelligence models through Metal in ways that unlock new rendering, simulation, or real-time content generation techniques?
Are there opportunities for developers to use Apple Intelligence models through Metal in ways that unlock new rendering, simulation, or real-time content generation techniques?
Metal provides several ways for you to use machine learning in your apps. There's TensorOps and the Metal 4 machine learning encoder introduced in WWDC25's Combine Metal 4 machine learning and graphics talk. This year's Build real-time neural rendering pipelines with Metal talk gives some more details about neural rendering applications and the Optimize custom machine learning operations with Metal tensors talk shows how to extend CoreAI with your own custom ops.