Compared to MLX - we've done some rough benchmarking and we are outperforming any of the MLX-based engines we've compared to so far. Going to do more in depth benchmarking and release it soon.