Superelement-based model order reduction offers a promising route for reducing the computational effort of thermomechanical reliability simulations, but its applicability depends on how accurately local solder joint responses are preserved. This work proposes a comparative evaluation methodology for assessing the agreement between full order and superelement-containing finite element models using a QFN package as demonstrator. The proposed evaluation methodology enables geometrically resolved comparisons throughout the load history. A quarter thermo-mechanical finite element model was established, the package region was reduced by component mode synthesis method, and both models were subjected to two thermal cycling conditions, i.e.,-40 to 85 °C (TC85) and-40 to 125 °C (TC125). The accumulated equivalent plastic strain in the solder joints was compared using four complementary methods: global error magnitude, global relative error, peak-value deviation, and spatial distribution analysis. For TC85, the superelement model reproduces both global and local solder joint responses with good accuracy, whereas for TC125 the deviations increase markedly once temperature-dependent material changes in the mold compound become relevant. The proposed framework therefore identifies both the accuracy and the temperature range of applicability of a given superelement implementation for solder joint reliability assessment.