The improvement in rolling contact fatigue life is a key subject to increase the reliability of bearings. It has been well known that the fatigue properties are significantly affected by nonmetallic inclusions in steel. The rolling contact fatigue life may be divided into two types, according to the type of reliability requirements. One is “L10 life,” which represents general bearing performance. The other is “accidental short life,” where a bearing prematurely fails in its service period and its calculated life is hardly met. Since the probabilities of these failures are significantly different, the distribution densities of the corresponding life-limiting inclusions are also expected to be widely different. When nonmetallic inclusions are evaluated as an index of fatigue life, therefore, testing volume should be appropriately determined according to the concerned type of life. From this point of view, it is effective for the reliability assessment of bearing steel to combine the evaluations of large microscopic inclusions by statistics of extreme value and macroscopic inclusions in larger volume by a high-frequency ultrasonic test. It was found that the L10 life determined by thrust-type rolling-contact fatigue tests was well correlated with maximum micro-inclusion diameter predicted by the statistics of extreme value, where both oxide and sulfide played as life-limiting inclusions. Fifteen MHz ultrasonic testing was employed to evaluate the macroscopic inclusions. The combination of these two procedures was able to demonstrate the difference in reliability clearly, even when other cleanliness indices could not characterize the concerned materials.