Skip to content
RAG Repo

Open Molecular Crystals 2025 (OMC25)

Open Molecular Crystals 2025 (OMC25) is a molecular-crystal dataset released by Meta's FAIR Chemistry team. It holds more than 27 million structures drawn from dispersion-inclusive DFT relaxation trajectories, with unit cells of up to 300 atoms and coverage of 12 elements. The crystals were generated with Genarris from roughly 50,000 organic molecules sampled from the OE62 set, starting from about 230,000 randomly generated candidates, and relaxed with the PBE functional in VASP. Each configuration carries lattice parameters, atomic positions, energies, forces and stress.

Access is through the Hugging Face hub, where the data is served as ASE-compatible LMDB files and split into training and validation sets. The repository is gated behind a contact-sharing agreement that asks for your full legal name, date of birth and organisation before download, and the companion fairchem library handles loading and training. Meta also published a baseline eSEN machine-learning interatomic potential trained on the full dataset.

For RAG and AI work this is a specialist scientific resource rather than a text corpus. It is best used to train or evaluate machine-learning interatomic potentials and foundation models for crystal structure prediction and organic solid-state chemistry, or to feed verifiable computed properties into a chemistry knowledge graph or structured retrieval layer rather than chunked prose.

On the licence, the dataset is released under CC BY 4.0, so commercial use is permitted with attribution, but there are real caveats. Access is gated and geographically restricted, being unavailable in comprehensively sanctioned jurisdictions and in China, Russia and Belarus, and the released model checkpoints ship under the separate FAIR Chemistry License v1 rather than CC BY, so do not assume the weights share the open data terms. Among our sibling entries, OMat24 covers inorganic bulk materials and OMol25 covers isolated molecules, while OMC25 fills the molecular-crystal gap between them.

chemistrymolecular-crystalsdftmachine-learning-potentialsmeta-fairscientific

Related sources