Core ML · 6 models available
Run a decision model on device with Core ML
Core ML puts the model on the Neural Engine, so the decision happens on the phone or the laptop and nothing leaves it. The strongest privacy story in the category.
How to run it
Download the .mlpackage
Core ML repositories ship a compiled package rather than raw weights.
Load it with coremltools or Swift
On device, the Neural Engine handles the forward pass and the model never touches the network.
pip install coremltoolsThis is the strongest privacy position available
If the decision happens on the phone, there is no payload to log, no region to choose and no processor agreement to sign. For regulated inputs that can matter more than accuracy.
Models available in Core ML
Every entry in the directory that ships this format, read from its repository metadata on 20 September 2026. None has been benchmarked by us.
laya-multilingual-coreml
A Core ML build of laya-multilingual, for running typed decisions on on-device Apple runtimes instead of a hosted API.
laya-coreml
A Core ML build of laya, for running typed decisions on on-device Apple runtimes instead of a hosted API.
laya-multilingual-coreml-ane
A Core ML build of laya-multilingual, for running typed decisions on on-device Apple runtimes instead of a hosted API.
laya-multilingual-coreml-ane-w8
A Core ML build of laya-multilingual, for running typed decisions on on-device Apple runtimes instead of a hosted API.
laya-multilingual-coreml-snake
A Core ML build of laya-multilingual, for running typed decisions on on-device Apple runtimes instead of a hosted API.
laya-typed-decisions-coreml
A Core ML build of laya-typed-decisions, for running typed decisions on on-device Apple runtimes instead of a hosted API.
Also browsable as a facet of the directory: Core ML builds.
Other ways to run one
GGUF
GGUF is the format that made local inference ordinary. For decision models it matters more than usual, because the whole argument for the category is cost per call.
MLX
MLX uses unified memory on M-series chips. For a model in the few-hundred-megabyte range this is the fastest local option on a Mac, and it needs no GPU of its own.
ONNX
An ONNX export loads without PyTorch. That matters for slim containers, for edge devices, and for running a decision inside a browser.
Compatible servers
16 open projects that reimplement the decision API, if you want the interface rather than just the weights.
Find out what your data actually supports
We turn your historical decisions into a frozen test set, then tell you which model, which threshold, and how much of it you can safely automate.