Ferrovanadium is essential for the production of many alloy steels. It is made by the aluminothermic reduction of vanadium oxides, together with scrap steel and burnt lime as a flux at very high temperatures. In this work, the theory of aluminothermic reduction is discussed, and then a computational thermodynamics model is described and its input parameters are discussed and justified. The model predicted very satisfactorily the composition of both the ferrovanadium and waste slag from a heat at the Windimurra smelter in Western Australia. Moreover, the modelling revealed that published data on the activity coefficient of VO1.5 in slag were likely to be seriously in error. It was shown that increasing the recovery of vanadium to the ferroalloy was accompanied by a rapid increase in its aluminium content, which is limited by commercial specifications. The limit on the silicon content of ferrovanadium can only be met by limiting the input of SiO2 to the furnace. The model provides a useful basis for assessing process improvements.