Broadening access to Skala creates a faster path to predictive DFT

Microsoft Research reports that its Skala 1.1 deep‑learning functional, trained on two and a half times more data than the original, now delivers gold‑medal accuracy on the GMTKN55 benchmark while costing as much as a traditional meta‑GGA. The model outperforms the most expensive global hybrid functionals across thermochemistry, reaction kinetics and non‑covalent interactions. Skala 1.1 has been rolled out in the popular electronic‑structure packages CP2K, Psi4, FHI‑aims, ORCA and VASP, and a living benchmark will continuously track performance as future releases arrive. Density functional theory, the workhorse of computational chemistry, is now more predictive and accessible for everyday scientific and industrial workflows.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton