UMA Release Changelog - FAIR Chemistry Documentation

This page documents the release history of UMA models, including new features, improvements, and bug fixes.

Library compatibility — UMA 1.0 deprecation¶

UMA 1.0 checkpoints are no longer supported (their MoE include_self behavior diverges from later releases); loading one raises a RuntimeError. To use a 1.0 checkpoint, pip install 'fairchem-core<=2.21.0'.

Loading now requires a model_id to be defined. UMA 1.1 (which ships without one) is auto-tagged model_id = "UMA-1.1" in memory; UMA 1.2+ already carry model_id; a checkpoint with neither model_id nor backbone.model_version is rejected. When training, set model_id on the model config (e.g. model_id: UMA-1.2.1).

UMA 1.2¶

Model Highlights¶

🔋 Better Charge Handling

Leading to improvements across molecular systems:

~50% relative improvement on biomolecules vs UMA-S-1.1
~30% relative improvement on electrolytes vs UMA-S-1.1

📊 Expanded Data Coverage

Now ~520M total DFT calculations including:

OC22 (oxide catalysts)
OC25 (electrolyte/inorganic interfaces)
Expanded OMol25 data including polymers (OPoly26)

⚡ Faster Inference

~50% speedup for UMA-S with turbo mode compared to previous releases

🔧 Bug Fixes & Stability

Numerical and stability improvements for Hessians and phonons
More accurate diatomics and ionization potentials

Available Tasks (7 total)¶

Task	Dataset	Domain	Status
oc20	OC20	Heterogeneous Catalysis	No update
omat	OMat24	Inorganic Materials	📊 Updated — dimer data
omol	OMol-1.0	Organic Molecules	📊 Updated — polymers (OPoly26), ionization potentials, small molecule clusters, dimers
odac	ODAC23	MOFs for Direct Air Capture	No update
omc	OMC25	Molecular Crystals	No update
oc25	OC25	Electrolyte/Inorganic Interfaces	🆕 Added to UMA
oc22	OC22	Oxide Catalysts	🆕 Added to UMA

Large-Scale Inference & Molecular Dynamics¶

🚀 Large-Scale MD Simulations

Run MD with ASE and LAMMPS on large-scale systems with ns/day speeds:

Battle tested up to hundreds of GPUs
Systems up to 1M atoms
UMA handles all parallelism—no need to manually configure LAMMPS, Kokkos, MPI, CUDA, etc.

📦 Batched Simulations

Client–server Ray framework enables batched simulations:

Structural relaxations
Molecular dynamics
Up to 4x speedup over serial calculations for batches of small systems on a single H100 GPU

UMA 1.1¶

Available Tasks (5 total)¶

Task	Dataset	Domain	Status
oc20	OC20	Heterogeneous Catalysis	No update
omat	OMat24	Inorganic Materials	No update
omol	OMol25	Organic Molecules	No update
odac	ODAC23	MOFs for Direct Air Capture	No update
omc	OMC25	Molecular Crystals	No update

UMA 1.0¶

Available Tasks (5 total)¶

Task	Dataset	Domain	Status
oc20	OC20	Heterogeneous Catalysis	🆕 New
omat	OMat24	Inorganic Materials	🆕 New
omol	OMol25	Organic Molecules	🆕 New
odac	ODAC23	MOFs for Direct Air Capture	🆕 New
omc	OMC25	Molecular Crystals	🆕 New

Key Features¶

Mixture-of-Linear-Experts (MoLE) architecture for high parameter count with fast inference
Equivariant GNN preserving energy conservation
Multi-domain training on 500M+ DFT calculations
ASE Calculator integration for seamless workflow adoption