Reverse Monte Carlo (RMC) methods [1] were originally developed to model the structure of liquids and amorphous materials, although there have been some applications to studies of highly disordered crystals. Recently we have reported a new program, RMCPOW [2-5], for modelling the atomic and magnetic structures of crystalline materials using powder diffraction data. In this method both Bragg and diffuse scattering are calculated for a model, consisting typically of hundreds of unit cells (thousands of atoms), and compared to experiment. Using a direct calculation of the scattering cross-section the experimental resolution can be taken into account and Fourier transform problems (as in the standard RMC approach) can be avoided. In addition magnetic structures with both long and short-range order can be modelled. Because RMCPOW can model both long range crystalline order and any local atomic correlations due to disorder it is a powerful complement to Rietveld refinement. Here we will present the RMCPOW method and recent modifications of the code. As one example of the interesting prospects for this new code we show the possibility to solve long-range order magnetic structures starting from random magnetic configurations, using as an illustration some results on the magnetic structure of Cr2O3.