A method for evaluating the reliability of steel in terms of inclusions is characterized in that the inclusions of approximately 100 μm or less in maximum inclusion size are evaluated by microscopy combined with the extreme-value statistical analysis; the inclusions of approximately 100 μm or greater in maximum inclusion size are evaluated by the ultrasonic flaw detection testing performed at a frequency of 5 to 25 MHz; and the reliability of the steel is evaluated based on the results of the extreme-value statistical analysis applied to the results of the microscopy and the results of the ultrasonic flaw detection testing combined.