Machine learning models for materials science and quantum chemistry.
State-of-the-art universal interatomic potentials for molecules, materials, and catalysts — built by the Meta FAIR Chemistry team.
pip install fairchem-coreApplication Domains¶
Surface reactions, adsorption energies, and catalyst design.
Bulk materials, phonons, and elastic properties.
Molecular conformers and electronic properties.
Packed molecular arrangements in crystal structures.
Metal-organic frameworks for CO₂ capture.
Get Started¶
Build your first calculation in minutes
Learn about task selection and model inputs
Set up FAIRChem and HuggingFace access
Hands-on tutorials for common workflows
Inference, training, fine-tuning, and more
Video presentations from the team
The UMA Model¶
Read about our latest release: the Universal Machine-learning for Atomistic systems (UMA) model.
UMA is trained on 500M+ DFT calculations across molecules, materials, and catalysts — achieving state-of-the-art accuracy with energy conservation and fast inference.