One of the most important issues that hinder the widespread use of friction stir (FS) processing, an effective microstructural modification technique, is the lack of accurate predictive tools that enable the selection of suitable processing parameters to obtain the desired grain structure. In this study, a model that is capable of predicting the resulting average grain size of a FS-processed material from process parameters is presented. The proposed model accounts for both dynamic recrystallization and grain growth. Several AZ31 magnesium samples were FS processed in different combinations of rotational and translational speeds. The thermal fields and resulting average grain size were measured, and the effective strain rates were approximated analytically. The results show that the proposed model is capable of predicting the resulting grain size of FS-processed materials.