The problem of uniqueness and representativeness of steel frame fire resistance assessment is considered in this paper. The thesis, that the selection of analysis method determines the result in both qualitative and quantitative terms is given scrutiny. It is also shown, that the differences between computed values may be significant. The selection of an appropriate computational model for an analysis of this type seems to be especially important, as the possible overestimation of the fire resistance determined during computation is equivalent to an unjustified optimism of the user with respect to the safety level warranted. In the considerations presented here the critical temperature determined for the whole bearing structure is considered as the measure of sought resistance. The determined temperature is associated with the bearing structure reaching the bearing capacity limit state subject to fire conditions, treated as accidental design situation. Two alternative computational methods have been applied during calculations: the first one – classical, based on 1st order statics and using the buckling length concept for members of the considered frame, and the second one – taking account of 2nd order phenomena via simple amplification of the horizontal loads applied to the frame. Special attention has been paid to the influence exerted on the final fire resistance of the considered structure by the real joint rigidity, decreasing with increasing temperature of the structural members. The obtained results differ not only in the value of determined temperature but also in the indicated location of the weakest frame component, determining its safety.