MLX from Apple, ported to Rust & Node.js, for super-fast local inference - View it on GitHub