I. Models

Matra-Genoa

Reference 10.1038/s41524-025-01940-8

Lead developer Pierre-Paul De Breuck

Our crystal generative model.

TUGA-SPComing soon

Lead developer Pierre-Paul De Breuck

Structure–property relations predicted directly from a crystal graph.

TUGA-FFComing soon

Lead developer Pierre-Paul De Breuck

A machine-learned interatomic potential for fast, accurate simulations.

Alexandria's reach goes beyond the models we build ourselves: its data has become a standard training ingredient for machine-learned interatomic potentials across the field. Data drawn from Alexandria underpins the training of essentially every top-ranked model on the Matbench Discovery leaderboard, the community benchmark for predicting materials stability.