UMA: A Family of Universal Models for Atoms

Welcome to UMA! This repo contains the model checkpoints for the UMA release: code paper

Overview

UMA is a large mixture-of-linear-experts graph network model trained on billions of atoms across five open-science simulation datasets released by the FAIR Chemistry team over the past 5 years. To explore the model's capabilities check out our educational demo.

Model checkpoints

Checkpoints are provided for UMA models (>=1.0), trained on the entirety of OC20, ODAC23, OMat24, OMC25, and OMol25 datasets.

Name	Checkpoint	Checksum (MD5)
uma-s-1.1	uma-s-1p1.pt	36a2f071350be0ee4c15e7ebdd16dde1
uma-s-1.2	uma-s-1p2.pt	26ac47f57e7d68af9f031077cdc2cbe9
uma-m-1.1	uma-m-1p1.pt	8936aecf2eb101089af85934d3e881d6

Archived checkpoints

uma-s-1 has a known extensivity bug so please use uma-s-1.1 or uma-s-1.2. This checkpoint remains available for reference purposes.

Name	Checkpoint	Checksum (MD5)
uma-s-1	uma-s-1.pt	dc9964d66d54746652a352f74ead19b6

Quick Start

The easiest way to use UMA is via the FAIRChemCalculator. You can find details about using the model, along with other utilities including tutorials and documentation in the official fairchem repo.

Support

If you run into any issues feel free to post your questions or comments on Github Issues.

License

Models are made accessible for commerical and non-commerical use under a permissive license found here.

UMA is available via HuggingFace globally, except in comprehensively sanctioned jurisdictions, and in China, Russia, and Belarus.

Citation

If you use this work, please cite:

@misc{wood2025umafamilyuniversalmodels,
      title={UMA: A Family of Universal Models for Atoms}, 
      author={Brandon M. Wood and Misko Dzamba and Xiang Fu and Meng Gao and Muhammed Shuaibi and Luis Barroso-Luque and Kareem Abdelmaqsoud and Vahe Gharakhanyan and John R. Kitchin and Daniel S. Levine and Kyle Michel and Anuroop Sriram and Taco Cohen and Abhishek Das and Ammar Rizvi and Sushree Jagriti Sahoo and Zachary W. Ulissi and C. Lawrence Zitnick},
      year={2025},
      eprint={2506.23971},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2506.23971}, 
}