whats new with core ml

At WWDC 2025, Apple introduced several updates and new features for Core ML, which is the framework used for deploying machine learning models on Apple devices. Here are some of the highlights:

1. Optimized Execution: Core ML optimizes execution across the CPU, GPU, and neural engine, making it efficient for a wide range of machine learning tasks such as segmentation, classification, and generative AI. This optimization ensures that models run efficiently on device, leveraging all available compute resources.

2. Integration with Metal 4: For applications requiring tight integration with the GPU, Core ML models can be used with Metal 4. This allows for advanced graphics and machine learning tasks, such as upscaling images and executing neural networks inside shaders, enhancing visual fidelity and performance in graphics-intensive applications.

3. BNNS Graph Enhancements: The BNNS (Basic Neural Network Subroutines) Graph API has been updated with a new graph builder, allowing developers to create graphs of operations. This enables the writing of pre and post-processing routines or small machine learning models to run in real-time on the CPU.

4. Model Visualization in Xcode: Xcode now provides a new view to visualize the structure of the full model architecture, helping developers understand and debug their models more effectively. This includes insights into prediction latency, load times, and operation support.

5. CoreML Tools: CoreML Tools offers utilities for transforming trained models into the Core ML format, with optimizations for on-device execution. It includes techniques for model compression to improve inference performance in terms of memory, power, and latency.

For more detailed information, you can refer to the session Discover machine learning & AI frameworks on Apple platforms (13:53).