Dispatch
Microsoft's Skala 1.1 DFT functional hits 2.8 kcal/mol on GMTKN55 and gold-medals 32 of 55 categories
Microsoft Research shipped Skala 1.1, a deep-learning exchange-correlation functional trained on 2.5x more data than 1.0, with added coverage of electron affinities and noncovalent clusters. It records a weighted average error of 2.8 kcal/mol on GMTKN55 and ranks first in 32 of that benchmark's 55 categories, beating expensive global hybrid functionals while keeping semi-local cost. It is integrated in CP2K, with Psi4, FHI-aims, ORCA and VASP in progress, plus an open-source community edition on PySCF with ASE support.
↳ Follow the thread