In recent years, low-cost silicon for photovoltaic applications has trended toward smaller grain sizes with lower dislocation densities. These changes, coupled with advanced device architectures like heterojunction and passivated emitter and rear cells, have enabled higher efficiencies, including new world records. To establish material requirements for further improvement, we model dislocations with different recombination strengths in Sentaurus Device using mid-gap defects spatially localized at the nanoscale in three dimensions. This approach accounts for the charging of dislocations and differentiates between dislocations and grain boundaries or metal impurities. Simulations that model dislocations as macroscopic variations in effective bulk lifetime or diode parameters do not address these issues. We validate our model by simulating the world record multicrystalline silicon solar cell. We find that three parameters influence cell efficiency: the area fraction of the cell containing dislocations, the dislocation density within defect clusters, and the concentration of recombination centers at these dislocations. We provide targets for these parameters and show how our results can be used to determine the potential gains from reducing dislocations in experimental devices and materials. We further predict that device architectures that lead to higher injection, such as silicon heterojunction cells, are more robust to the presence of dislocations. (C) 2016 Elsevier B.V. All rights reserved.