[inkyoto]

FP16 is supported in M1 GPU's and Neural Engines through the CoreML framework. From https://coremltools.readme.io/docs/typed-execution :

> The Core ML runtime dynamically partitions the network graph into sections for the Apple Neural Engine (ANE), GPU, and CPU, and each unit executes its section of the network using its native type to maximize its performance and the model’s overall performance. The GPU and ANE use float 16 precision, and the CPU uses float 32.

Also, this exploration (https://tlkh.dev/benchmarking-the-apple-m1-max#heading-neura...) reports the 5.1-5.3 TFLOPS FP16 ballpark performance.