Microsoft Research has released Skala-1.1, an updated density functional theory (DFT) model that delivers substantially improved performance across key molecular simulation challenges including main-group thermochemistry, reaction kinetics, and molecular structure prediction. The new version achieves a weighted average error of 2.8 kcal/mol on GMTKN55, a widely used benchmark suite of 55 chemistry categories, surpassing leading global hybrid functionals while retaining the efficiency of a semi-local functional.
To broaden accessibility, Skala is being integrated into multiple leading electronic-structure software packages: it is already available in CP2K and is being integrated into Psi4, FHI-aims, ORCA, and VASP. The team is also introducing a living benchmark to track computational performance across successive optimized Skala releases, providing transparent reference points for implementations across software packages and hardware platforms.
Unlike the traditional "functional zoo" where new functionals accumulate without replacing older ones, Skala follows a philosophy where each release is designed to supersede the previous version as new data, model architectures, and training strategies become available. Beyond energies, Skala-1.1 provides highly accurate electron densities, dipole moments, and molecular geometries, enabling broader scientific and industrial workflows in chemistry, materials science, catalysis, energy technologies, and drug discovery.