Run a server
Pick MLX, llama.cpp, oMLX, or MTPLX and follow the shortest path from installation to a running local model.
Local AI tooling
A practical starting point for running open-source models on Apple Silicon.
Local models come with a lot of choices: models, runtimes, formats, quantizations, context sizes, and configuration. You don't need to understand all of that before trying one.
I keep the runtime guides, conceptual notes, and model examples together so the first attempt stays practical. The point is to have somewhere useful to go when the next choice gets confusing.
Pick MLX, llama.cpp, oMLX, or MTPLX and follow the shortest path from installation to a running local model.
New to local AI? Start with plain-language explanations of models, runtimes, quantization, GGUF, context, memory, and the terminology you'll encounter.
See models I've actually used, including the artifact, runtime, and machine where they gave me a useful experience.
Start here
Clone the repository to keep the commands and implementation details close at hand. The project focuses on Apple Silicon and stays a practical reference rather than a packaged inference runtime.
git clone https://github.com/joacod/local-ai.git
cd local-aiUse it
Choose the path that matches your starting point:
Local AI isn't a benchmark suite or an exhaustive model catalog. The model and machine notes document combinations I've actually used so they can give someone with similar hardware a useful place to start.