In this work we study and compare hybrid versions of two different metaheuristics to solve a real-world frequency assignment problem (FAP) in GSM networks. We have used a precise mathematical formulation, developed in published work, in which the frequency plans are evaluated using accurate interference information coming from a real GSM network. Therefore, we focus on solving FAP problem for a realistic-sized, real-world GSM network using a hybrid version of Scatter Search and Differential Evolution algorithms. We have analyzed and fixed both approaches to the FAP problem and, after a detailed statistical study, the obtained results prove that our hybrid approaches compute accurate frequency plans for real-world instances in an optimum way. In fact, our results surpass all the results previously published in the literature.