GPCRs are key components of signal transduction pathways and are important drug targets. Recently determined GPCR structures provide opportunities for advancements in GPCR modeling. This review focuses on the choice of experimental templates, the treatment of extracellular loops and the description of ligand-binding sites in GPCR modeling. Four important conclusions are reached in this review: (i) multi-template models may produce better structures than single-template models, although inferior models may also be generated by multi-template approaches, warranting the development and application of improved model assessment methods; (ii) cautious incorporation of knowledge-based constraints can improve the quality of models and docking; (iii) molecular dynamics simulations account for structural features not observed in X-ray structures and may refine docking poses; and (iv) while progress in de novo methods for long loop prediction is ongoing, loopless models provide a practical alternative for docking and virtual screening applications.