The possibility of reprogramming human somatic cells to pluripotency has opened unprecedented opportunities for creating genuinely human experimental models of disease. Inborn errors of metabolism (IEMs) constitute a greatly heterogeneous class of diseases that appear, in principle, especially suited to be modeled by iPSC-based technology. Indeed, dozens of IEMs have already been modeled to some extent using patient-specific iPSCs. Here, we review the advantages and disadvantages of iPSC-based disease modeling in the context of IEMs, as well as particular challenges associated to this approach, together with solutions researchers have proposed to tackle them. We have structured this review around six lessons that we have learnt from those previous modeling efforts, and that we believe should be carefully considered by researchers wishing to embark in future iPSC-based models of IEMs.