A Part I accompanying paper clearly showed that for a given level of strain range (Δvar epsilon), the number of cycles required to achieve crack initiation is significantly lower in biaxial thermal fatigue than in uniaxial isothermal fatigue. Such discrepancy does not come from a thermal effect, as it is emphasized by a comparison between the uniaxial data, coming either from thermomechanical fatigue tests, or from isothermal low cycle fatigue tests. A detrimental effect of multiaxial loading on fatigue life is noticed in many cases as reported in the literature. In this frame, the goal of this paper is to propose criteria in order to take into account the reduction of fatigue life resulting from the multiaxiality. Among all the examined multiaxial criteria, a criterion based on strain proposed by Zamrik, and a new criterion based on energy proposed by “Ecole Polytechnique” give the best estimations. Furthermore, application of the Zamrik's criterion using the RCC–MR method is very promising for the design engineer. Indeed, these two criteria exhibit a notable potential, since they are well adapted to all the available thermal fatigue experimental data, despite significant differences between approaches and processes developed by both accessible facilities. However, such investigation must be continued with achievement of additional thermal fatigue tests (for higher number of cycles...), and isothermal biaxial fatigue tests.