A collaborative study supported by the European Community COST 50 program has been conducted on life prediction methods under high-temperature low-cycle fatigue and thermal fatigue cycling in the case of MAR-M509, a cast cobalt-base superalloy. Low-cycle fatigue tests were carried out by the Ecole des Mines de Paris at 900°C in air under longitudinal strain control and included sawtooth triangular shape cycles at three strain rates (frequency range 10⁻³ to 20 Hz), tensile strain hold cycles, and compressive strain hold cycles. Life to initiation data deduced from a-c potential drop measurements were used to evaluate three types of predictive models: strainrange partitioning, studied by BBC-Baden; creep damage models, studied by the University of Leoben; and an oxidation-fatigue interaction model, studied by the Ecole des Mines de Paris. All the methods have been found to correlate experimental data within a factor of three. However, they range from simple correlating methods with adjustable parameters to be fitted to fully predictive methods with no fitting parameter. Thermal fatigue data on wedge-type specimens (i.e., components instead of laboratory specimens) have been used to check the prediction capability of the various methods and have shown the superiority of oxidation-fatigue interaction models for this material.