Phosphate hydrolysis is ubiquitous in biology. However, despite intensive research on this class of reactions, the precise nature of the reaction mechanism remains controversial. Herein, we have examined the hydrolysis of three homologous phosphate diesters. The solvation free energy was simulated by means of either an implicit solvation model (COSMO), hybrid quantum mechanical/molecular mechanical free energy perturbation (QM/MM-FEP) or a mixed solvation model in which N water molecules were explicitly included in the ab initio description of the reacting system (where N=1-3), with the remainder of the solvent being implicitly modelled as a continuum. Here, both COSMO and QM/MM-FEP reproduce DeltaG(obs) within an error of about 1 kcal mol(-1). However, we demonstrate that in order to obtain any kind of reliable results from a mixed model, it is essential to carefully select the explicit water molecules from short QM/MM runs that act as a model for the true infinite system. Additionally, the mixed models tend to be increasingly unstable and miss larger entropic contributions as more explicit water molecules are placed into the system. Thus, our analysis indicates that this approach provides an unreliable way for modelling phosphate hydrolysis in solution.