They do work with llama.cpp and MLX automatically.