A thermodynamic model is formulated for (Ca,Na)2(Mg,Fe2+,Al,Fe3+)T1 (Al,Fe3+,Si ) 2 T2 O7 melilites. It employs the compositional vertices: akermanite (Ca2MgSi2O7, 1), gehlenite (Ca2Al2SiO7, 2), iron akermanite (Ca2Fe2+Si2O7, 3), ferrigehlenite (Ca2 Fe 2 3 + SiO7, 4), sodium melilite (NaCaAlSi2O7, 5), and the convergent ordering variables: s = X Al 3 + T2a – X Al 3+ T2b and t = X Fe 3 + T2a – X Fe 3 + T2b to describe the distribution of Al3+, Fe3+ and Si4+ between T2 subsites T2a and T2b. It is calibrated for akermanite–gehlenite melilites based on the calorimetric data of Charlu and others (1981), the assumption that the synthetic samples of Charlu and others approached “equilibrium” states of Al-Si tetrahedral ordering at 970 K, and analogy with the Al2(MgSi) − 1 substitution in CaMgSi2O6 – CaMg1/2 Ti 1 / 2 AlSiO6 – CaAl2SiO6 fassaites (for example, Sack and Ghiorso, 2017). In this model gehlenite has a disordered Al-Si distribution on T2 sites above 1443 K (1170 °C), consistent with the crystallographic data on c/a ratios of lattice parameters as a function of annealing temperature (Woodhead and Waldbaum, 1974) and the high-temperature heat capacities inferred from drop calorimetric data (Pankratz and Kelley,1964). However, above this critical temperature a partially ordered Al-Si distribution persists between T2a and T2b sites in akermanite – gehlenite solid solutions with intermediate X2 (for example, 0.19 X 2 Δ H ¯ f 298.15 o AK and Δ H ¯ f 298.15 o GEHL , consistent with the experimental brackets on decarbonation equilibria of Walter (1963), Hoschek (1974), and Shmulovich (1974), the thermodynamic model for akermanite-gehlenite melilites developed here, the thermodynamic properties of the other phases in these reactions tabulated by Berman (1988), and the revised estimates for C ¯ p and S ¯ 298.15 o of diopside of Sack and Ghiorso (2017), are roughly 1 and 3 (kJ/gfw) more positive than those estimated by Berman (1988). More positive standard enthalpies of formation of both endmembers, together with a decrease in the vibrational heat capacity of gehlenite and less negative deviations from ideal mixing compared with previous calibrations, all contribute to reducing the stability of melilites in this model. Together these effects will decrease the predicted temperature of condensation of melilite from nebular vapors, bringing calculated temperatures of melilite condensation into closer alignment with those of MgAl2O4 spinel than the 80 to 100 K separating their appearances in previous calculations (for example, Yoneda and Grossman, 1995; Petaev and Wood,1998; Ebel and Grossman,2000). These effects, together with a possible increase in spinel stability due to non-negligible solubility of Al8/3O4 alumina component, may allow equilibrium models to match the observed condensation sequence of spinel before melilite in calcium-aluminum inclusions (CAIs) in carbonaceous chondrites, without the need to invoke kinetic effects.