Life predictions given by three creep-fatigue damage models, namely the‘continuous damage model’, the‘damage rate approach’and the‘time and cycle fractions summation’method, were evaluated for over 200 high temperature fatigue tests reported in the literature. These tests, which were conducted on AISI 304 and 316 stainless steels, include several different multiaxial and time-variable loading conditions (in-phase and out-of-phase biaxial stress, various strain waveforms, complex loading histories, etc.) whose influence on the distribution of the ratio between predicted and observed cycles to failure was analysed. Results obtained by the three models are discussed, emphasizing differences in accuracy and cases in which the creep-fatigue behaviour does not appear to be satisfactorily predicted.