fairchem by the FAIR Chemistry team

fairchem is the FAIR Chemistry's centralized repository of all its data, models, demos, and application efforts for materials science and quantum chemistry.

⚠️FAIRChem version 2 is a breaking change from version 1 and is not compatible with our previous pretrained models and code. If you want to use an older model or code from version 1 you will need to install version 1, as detailed here.

Caution

UMA models and legacy inorganic bulk models trained using OMat24 are trained with DFT and DFT+U total energy labels. These are not compatible with Materials Project calculations. If you are using UMA or models trained on OMat24 only for such calculations, you can find a OMat24 specific calculations of reference unary compounds and MP2020-style anion and GGA/GGA+U mixing corrections in the OMat24 Hugging Face repo. Do not use MP2020 corrections or use the MP references compounds when using OMat24 trained models. Additional care must be taken when computing energy differences, such as formation and energy above hull and comparing with calculations in the Materials Project since DFT pseudopotentials are different and magnetic ground states may differ as well.

Latest news

March 2026 - UMA-1.2 released! ~50% faster, ~40% more accurate on Open Molecules test set, and expanded data coverage for catalysts (oxides and interfaces), molecules, and polymers!

Oct 2025 - check out our seamless Multi-node, Multi-GPU and LAMMPs interfaces to run large scale dynamics!

Read our latest release post!

Read about the UMA model and OMol25 dataset release.

Try the demo!

If you want to explore model capabilities check out our educational demo

Installation

Although not required, we highly recommend installing using a package manager and virtualenv such as uv, it is much faster and better at resolving dependencies than standalone pip.

Install fairchem-core using pip

pip install fairchem-core

If you want to contribute or make modifications to the code, clone the repo and install in edit mode

git clone git@github.com:facebookresearch/fairchem.git
pip install -e fairchem/packages/fairchem-core[dev]

Quick Start

The easiest way to use pretrained models is via the ASEFAIRChemCalculator. A single uma model can be used for a wide range of applications in chemistry and materials science by picking the appropriate task name for domain specific prediction.

Instantiate a calculator from a pretrained model

Make sure you have a Hugging Face account, have already applied for model access to the UMA model repository, and have logged in to Hugging Face using an access token. You can use the following to save an auth token,

huggingface-cli login

Models are referenced by their name, below are the currently supported models:

Model Name	Description
uma-s-1p2	Latest version of the UMA small model, fastest of the UMA models while still SOTA on most benchmarks (6.6M/290M active/total params)
uma-s-1p1	Early version of the UMA small model while still SOTA on most benchmarks (6.6M/150M active/total params)
uma-m-1p1	Best in class UMA model across all metrics, but slower and more memory intensive than uma-s (50M/1.4B active/total params)

Set the task for your application and calculate

• oc20: use this for catalysis
• oc22: use this for oxide catalysis (1p2 only)
• oc25: use this for (electro)catalysis (1p2 only)
• omat: use this for inorganic materials
• omol: use this for molecules+polymers
• odac: use this for MOFs
• omc: use this for molecular crystals

Relax an adsorbate on a catalytic surface,

fromase.buildimportfcc100, add_adsorbate, moleculefromase.optimizeimportLBFGSfromfairchem.coreimportpretrained_mlip, FAIRChemCalculatorpredictor=pretrained_mlip.get_predict_unit("uma-s-1p2", device="cuda")
calc=FAIRChemCalculator(predictor, task_name="oc20")
# Set up your system as an ASE atoms objectslab=fcc100("Cu", (3, 3, 3), vacuum=8, periodic=True)
adsorbate=molecule("CO")
add_adsorbate(slab, adsorbate, 2.0, "bridge")
slab.calc=calc# Set up LBFGS dynamics objectopt=LBFGS(slab)
opt.run(0.05, 100)

Relax an inorganic crystal,

fromase.buildimportbulkfromase.optimizeimportFIREfromase.filtersimportFrechetCellFilterfromfairchem.coreimportpretrained_mlip, FAIRChemCalculatorpredictor=pretrained_mlip.get_predict_unit("uma-s-1p2", device="cuda")
calc=FAIRChemCalculator(predictor, task_name="omat")
atoms=bulk("Fe")
atoms.calc=calcopt=FIRE(FrechetCellFilter(atoms))
opt.run(0.05, 100)

Run Molecular Dynamics (MD)

Note: pretrained_mlip.get_predict_unit() currently uses a seed to set the global state of the numpy RNG. In order to obtain different trajectories for different runs of the following code, we have to set a random seed as shown below:

fromaseimportunitsfromase.ioimportTrajectoryfromase.md.langevinimportLangevinfromase.buildimportmoleculefromfairchem.coreimportpretrained_mlip, FAIRChemCalculatorseed=np.random.randint(0, np.iinfo(np.int32).max, dtype=int)
predictor=pretrained_mlip.get_predict_unit("uma-s-1p2", device="cuda", seed=seed)
calc=FAIRChemCalculator(predictor, task_name="omol")
atoms=molecule("H2O")
atoms.calc=calcdyn=Langevin(
atoms,
timestep=0.1*units.fs,
temperature_K=400,
friction=0.001/units.fs,
)
trajectory=Trajectory("my_md.traj", "w", atoms)
dyn.attach(trajectory.write, interval=1)
dyn.run(steps=1000)

Calculate a spin gap,

fromase.buildimportmoleculefromfairchem.coreimportpretrained_mlip, FAIRChemCalculatorpredictor=pretrained_mlip.get_predict_unit("uma-s-1p2", device="cuda")
# singlet CH2singlet=molecule("CH2_s1A1d")
singlet.info.update({"spin": 1, "charge": 0})
singlet.calc=FAIRChemCalculator(predictor, task_name="omol")
# triplet CH2triplet=molecule("CH2_s3B1d")
triplet.info.update({"spin": 3, "charge": 0})
triplet.calc=FAIRChemCalculator(predictor, task_name="omol")
triplet.get_potential_energy() -singlet.get_potential_energy()

Multi-GPU Inference and LAMMPs

If you have multiple gpus (or multiple nodes), we handle all the parallelism for you under the hood by a single flag (workers=N). For example, you can run the following 8000 atom md simulation with ~10 qps (8x H100 GPU), ~10x faster than single-gpu inference! Current benchmarks show we can run uma-s @ ~1 ns/per day with 100k+ atoms systems in real MD scenarios (more on this to come!). This is also compatible with LAMMPs to perform large scale MD. See our docs for more details. This requires the Ray package to be installed and comes with the extras bundle.

pip install fairchem-core[extras]

fromaseimportunitsfromase.md.langevinimportLangevinfromfairchem.coreimportpretrained_mlip, FAIRChemCalculatorimporttimefromfairchem.core.datasets.common_structuresimportget_fcc_crystal_by_num_atomsseed=np.random.randint(0, np.iinfo(np.int32).max, dtype=int)
predictor=pretrained_mlip.get_predict_unit(
"uma-s-1p2", inference_settings="turbo", device="cuda", workers=8, seed=seed,
)
calc=FAIRChemCalculator(predictor, task_name="omat")
atoms=get_fcc_crystal_by_num_atoms(8000)
atoms.calc=calcdyn=Langevin(
atoms,
timestep=0.1*units.fs,
temperature_K=400,
friction=0.001/units.fs,
)
# warmup 10 stepsdyn.run(steps=10)
start_time=time.time()
dyn.attach(
lambda: print(
f"Step: {dyn.get_number_of_steps()}, E: {atoms.get_potential_energy():.3f} eV, "f"QPS: {dyn.get_number_of_steps()/(time.time()-start_time):.2f}"
),
interval=1,
)
dyn.run(steps=1000)

LICENSE

fairchem is available under a MIT License. Models/checkpoint licenses vary by application area.