MLX · 5 models available
Run a decision model on Apple Silicon
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.
How to run it
Install mlx-lm
Apple Silicon only. MLX shares memory between CPU and GPU, so there is no copy step and no separate VRAM budget.
pip install mlx-lmRun the model
Weights download from Hugging Face on first use.
mlx_lm.generate --model <owner>/<repo>Expect this to be the fastest local option on a Mac
For models in the few-hundred-megabyte range the whole thing sits in unified memory, which is exactly the size range most of this category lives in.
Models available in MLX
Every entry in the directory that ships this format, read from its repository metadata on 20 September 2026. None has been benchmarked by us.
Qwen-2.5-1B-RLCD
Qwen2.5-1.5B-Instruct fine-tuned to return typed decisions.
laya-mlx
laya fine-tuned to return typed decisions, 421M parameters.
laya-multilingual-mlx
laya-multilingual fine-tuned to return typed decisions, 322M parameters.
gemma-e2b-rlcd
A MLX build of gemma-4-E2B-it, for running typed decisions on Apple Silicon instead of a hosted API.
laya-typed-decisions-mlx
laya-typed-decisions fine-tuned to return typed decisions, 421M parameters.
Also browsable as a facet of the directory: MLX 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.
ONNX
An ONNX export loads without PyTorch. That matters for slim containers, for edge devices, and for running a decision inside a browser.
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.
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.