Inclusions play a crucial role for the fatigue properties of high strength steel, but to find thelargest inclusions by microscopy measurements large areas have to be examined. In this study ultrasonic gigacycle staircase fatigue testing has been used to find large inclusions in an H13 tool steel. The inclusions have been examined in SEM and their sizedistribution modeled using methods from extreme value statistics. The inclusion distribution obtained from the fatigue crack surfaces is compared to distributions acquired by microscopy study of cross sections as well as ultrasound immersion tank measurements and to the corresponding staircase fatigue data via the Murakami \sqrt{Area} model. It is shown that the fatigue method more effectively finds large inclusions than the other methods. It is also shown that the correlation between predictions of inclusion sizes by the \sqrt{Area} model from stress levels and fatigue initiating inclusions is weak for this material