Microsoft Research AI for Science has made its Skala AI model available through the open-source simulation platform CP2K, giving researchers access to a machine-learned exchange-correlation functional that can handle larger molecular systems with higher accuracy than conventional approaches. The integration, developed in collaboration with the CP2K team at the Center for Advanced Systems Understanding (CASUS) at Helmholtz-Zentrum Dresden-Rossendorf, went live in late August 2026 after a collaboration that began in early 2026.
Density functional theory (DFT) underpins much of modern computational chemistry and materials science, but it has a known weak point. "The Achilles' heel of DFT is the so-called exchange-correlation functional," said CASUS Director Prof. Thomas D. Kühne. The most accurate functionals require so much computation time that they can only be applied to systems with a small number of particles.
Skala takes a different route. Instead of adding another layer of mathematical formulas, Microsoft's team trained a neural network to learn how electron densities in different regions influence one another. That makes Skala one of the first AI-based exchange-correlation functionals available to researchers.
First results from the collaboration
The teams published a preprint in mid-August presenting initial results of the integration. "The results presented by Microsoft Research in 2025 were impressive, particularly the combination of accuracy and computational efficiency for certain DFT calculations," Kühne said. "We saw an opportunity to evaluate the approach within CP2K and better understand its applicability to problems relevant to our community."
Lead author Franz Pöschel of CASUS' Scientific Computing Core confirmed the model delivers in practice. "I can confirm that with Skala we've achieved a noticeable leap in the accuracy of our simulations for our specific test case," he said.
Interest in the integration has been building since Microsoft Research announced the plan at a conference in spring 2026. "Even though we were pleased by the interest, we felt a certain pressure to deliver," Pöschel said. "I'm therefore glad to confirm that Skala is now available within CP2K for simulations of molecular systems."
Why CP2K was the target
CP2K is a widely used platform for DFT calculations, particularly for dynamic simulations of large systems over long time periods. It can handle molecules, liquids, solids, and biological systems using quantum mechanical methods, and its algorithms can compute systems with thousands to tens of thousands of atoms. That makes it a natural fit for research into battery materials, catalysts, semiconductors, and proteins.
"To increase adoption, Skala should be available where the community already works," said Dr. Sebastian Ehlert, Senior Researcher at Microsoft Research AI for Science. "CP2K has been a cornerstone of computational chemistry research for many years, making it a natural priority for integration."
CP2K recently added features that support AI-based models, which can predict molecular energies and forces from training data generated by CP2K itself. This expands the time and length scales researchers can simulate.
Testing and what comes next
Before release, the two teams ran extensive tests on the integration. "We want to ensure that Skala delivers consistent accuracy and speed across different programs and settings," Ehlert said. "Together with the CP2K team, we created a set of integration tests to ensure that Skala delivers numerically correct results."
Skala is not a static model. Microsoft continues to improve it, and future releases will support periodic solids such as metals and semiconductors, as well as liquids. The CP2K integration is designed to make it easier for users to access the latest Skala version as updates arrive.
Why this matters for researchers
For computational chemists and materials scientists, the practical takeaway is that AI-based exchange-correlation functionals have moved from research demonstrations to production-ready tools in a mainstream simulation platform. The CP2K integration means researchers can test Skala on their own systems without building custom workflows or switching software. The published preprint also provides a reference point for benchmarking the model's accuracy against conventional functionals, which should help the community assess where Skala fits into their simulation pipelines.
Your membership also unlocks:
