It has been proved that rho, which can be obtained by Equation 4, is the maximal reliability. In the single factor model, by definition, a reliability omega can be derived from Equation 3. The reduction in reliability from rho to omega is solely caused by the loss of information when raw sum score is used. In addition, the Cronbach’s alpha is a lower bound of omega and equals to omega when the items load equally on the underlying factor. On the other hand, the reliability pi for IRT model, as depicted in Equation 10, is defined by the information function, and its value is no larger than rho. Thus the relationship between omega and pi was investigated in the study. After a series of simulations, it was found that omega is not necessarily greater than pi under all conditions. It could happen when the slope values vary a lot among items, or when response categories increase to a certain level.