Complex phenomena often require complex explanations. In simple cases, a single model can explain a phenomenon, but fully to explain a complex phenomenon, we often require systems of models. In this paper, we consider how to combine models that differ from each other along multiple lines, for instance, in the way they represent a phenomenon or the variables they include. We provide an interpretation of such situations in terms of structural realism. We focus on the conditions that allow scientists to compare different models so as to combine their results and obtain new knowledge claims.