Broadening access to Skala creates a faster path to predictive DFT
2026-08-21 00:00Models🔥 28.9 heat score
1sources
1days unfolding
28.9heat score
2mentions
SummaryAI generated
On August 20, 2026, Microsoft Research released Skala 1.1. Based on a training dataset that is 2.5 times larger than previous versions, this model significantly improved the accuracy in key areas such as thermochemistry, reaction kinetics, and molecular structure prediction. To expand its application range, Skala has been integrated into the CP2K software package and is being incorporated into several mainstream computational chemistry tools, including Psi4, FHI-aims, ORCA, and VASP, with the aim of providing high-precision prediction capabilities to research communities that rely on these tools. Additionally, Microsoft Research introduced a “live benchmark” mechanism to continuously track the performance optimization progress of each version of Skala, thereby accelerating the improvement of the model’s accuracy and efficiency.
Microsoft Research has released Skala 1.1. After increasing the amount of training data by 2.5 times compared to previous versions, this model has achieved significantly higher accuracy in key molecular simulation challenges such as thermochemistry, reaction kinetics, and molecular structure prediction. Skala has now been integrated into CP2K and is being incorporated into the Psi4, FHI-aims, ORCA, and VASP software packages, aiming to bring the predictive accuracy of next-generation density functional theory (DFT) to the research community that relies on these tools. Microsoft Research also introduced a “live benchmark” to track the optimization progress of computational performance in successive Skala versions, thereby accelerating improvements in accuracy and efficiency.