New activity–composition (a–x) relations for minerals commonly occurring in metapelites are presented for use with the internally consistent thermodynamic dataset of Holland & Powell (, Journal of Metamorphic Geology, 29, 333–383). The a–x relations include a broader consideration of Fe2O3 in minerals, changes to the formalism of several phases and order–disorder in all ferromagnesian minerals where Fe–Mg mixing occurs on multiple sites. The a–x relations for chlorite, biotite, garnet, chloritoid, staurolite, cordierite, orthopyroxene, muscovite, paragonite and margarite have been substantially reparameterized using the approach outlined in the companion paper in this issue. For the first time, the entire set of a–x relations for the common ferromagnesian minerals in metapelitic rocks is parameterized simultaneously, with attention paid to ensuring that they can be used together to calculate phase diagrams of geologically appropriate topology. The a–x relations developed are for use in the Na2O–CaO–K2O–FeO–MgO–Al2O3–SiO2–H2O–TiO2–O2 (NCKFMASHTO) system for both subsolidus and suprasolidus conditions. Petrogenetic grids in KFMASH and KFMASHTO are similar in topology to those produced with earlier end‐member datasets and a–x relations, but with some notable differences. In particular, in subsolidus equilibria, the FeO/(FeO + MgO) of garnet is now greater than in coexisting staurolite, bringing a number of key staurolite‐bearing equilibria into better agreement with inferences from field and petrographic observations. Furthermore, the addition of Fe3+ and Ti to a number of silicate phases allows more plausible equilibria to be calculated in relevant systems. Pseudosections calculated with the new a–x relations are also topologically similar to equivalent diagrams using earlier a–x relations, although with many low variance fields shifting in P–T space to somewhat lower pressure conditions.

A fully thermodynamic model for mafic melt in CaO–MgO–Al2O3–SiO2 (CMAS) has been calibrated, for calculation of melting equilibria in the pressure range 0–50 kbar. It is intended as a preliminary step towards a large‐system melt model, suitable for exploring melting, melt loss and crystallization processes in a wide range of natural rock compositions. Calibration was performed with attention to the model's behaviour in its compositional subsystems, as a rigorous test of model structure and parameterization. The model is consistent with the latest Holland & Powell thermodynamic data set, and can therefore be used to calculate phase relations in conjunction with the many solid‐phase activity–composition models written for the data set. Model calculations successfully reproduce experimental melting reactions in CMAS spinel lherzolite and garnet lherzolite assemblages, as well as sapphirine‐ and kyanite‐bearing assemblages, at moderate to high pressure. Thermodynamically sensitive features, such as thermal divides are also recovered. However, some changes to the model structure will be required before the model can describe the full range of mafic and ultramafic melt compositions known from experiment at low pressures.