what's the point of this benchmark if sonnet is working great at my tasks and mini can't solve my tasks?