The recent publication of new activity-composition models by Holland, Green and Powell (2018; Journal of Petrology 59: 881-900), with a melt model calibrated for source compositions ranging from peridotitic to granitic, opens the door to the modelling of multiple petrogenetic processes at supersolidus conditions in which the composition of the melt phase changes considerably, without having to change the melt model. This melt model is also the first one using the internally consistent thermodynamic databases published by T. Holland and R. Powell that contains TiO 2 and Fe 2 O 3 , further expanding the application of this model to more realistic geological scenarios. The accompanying mineral models are also the first in containing some minor elements, like TiO 2 in garnet and K 2 O in clinopyroxene. Consequently, it is relevant to test the applicability of these new models to a large P-T-X range of conditions before they can be used in full. Thermodynamic calculations made with the software Perple_X using these models were compared to experimental results, namely the modal proportions and the composition of the melt and several mineral solution phases. The experiments chosen for the comparative study covered a wide range of source compositions (from mafic to felsic), pressure (from 0.3 to 2.1 GPa), temperature (from 700 to 1,150°C) and total and added water content (structural water: 0.15-1.48 wt%; added water: 0-8 wt%; total water: 0.15%-8.15%). The results indicate that the extended melt model reproduces well the composition of the experimental melts, with an inverse correlation between component amount and fit: the best match is found for SiO 2 (−0.8% on average) and the worst match is found for those elements with the lowest amounts, TiO 2 and MgO (+241% and +235% on average, respectively; values indicate calculated minus experimental, times 100 and divided by experimental). The TiO 2 content in the melt model increases dramatically with increasing pressure, from +90% for P < 1.5 GPa to +593% for P > 1.5 GPa. No comparison was made on the Fe 2 O 3 content, as the published iron contents of the experimental melts were always reported as FeO t. In some cases, there is a substantial mismatch in the modal proportions between experiments and calculations, with the reactant phases less abundant and product phases more abundant in the calculations, an effect that is attributed to kinetic effects in the experiments and to the selected clinoamphibole model. Finally, the extended melt model was compared 994 | GARCÍA-ARIAS 1 | INTRODUCTION Thermodynamic calculations have become a powerful tool to study several aspects of igneous and metamorphic rocks, not only the P-T conditions at which a mineral assemblage equilibrated but also the fertility of crustal sources, open system behaviour and melt drainage during partial melting, alteration vectors in hydrothermally altered rocks, deformation and tectonics, and many more topics (e.g. Yakymchuk, 2017). A particular set of calculations are the isochemical phase diagrams (pseud...