Chemical zoning of garnet is often used to deduce P–T paths of rocks by inverse calculation. To validate the derived P–T paths, it is desired to establish a method to predict the chemical compositions of garnet theoretically. This study proposes a new forward calculation of the formation of Mg–Fe–Mn garnet from chlorite, which solves the non-linear simultaneous equations using nested iterative calculations. Growth of garnet consuming chlorite and quartz was modelled in a MnO–FeO–MgO–Al2O3–SiO2–H2O system, using the most recent thermodynamic data for the minerals. The prograde P–T history of the Sambagawa metamorphic belt, SW Japan, was modelled. To reproduce growth zoning, crystallized garnet was removed step by step from the system; perfect diffusion was assumed for chlorite. The proposed model derived the evolution of molar amounts and chemical compositions of Mg–Fe–Mn chlorite and garnet. It successfully reproduced the shape of the observed chemical profile of garnet, although the temperature condition was higher than general observations. The Mn content of the garnet core was generally high, and Mg/Fe ratio always started rising rapidly after Mn was depleted. Thermodynamic properties of minerals, initial chlorite composition, P–T path, H2O partial pressure, and Ca content in garnet were varied to test the behaviour of the system. The properties of Mn phases influenced only the chemical composition of the garnet core. The temperature range in which garnet grew depended on the H2O partial pressure or the Ca content in garnet.