You're better served using Apple's MLX if you want to run models locally.