The Open Molecules 2025 (OMol25) Dataset, Evaluations, and Models

TL;DR

OMol25 addresses a critical data bottleneck in molecular ML by delivering a massive, high-fidelity DFT dataset that spans biomolecules, metal complexes, and electrolytes, generated with uniform ωB97M-V/def2-TZVPD theory and diverse sampling. The authors accompany the data with baseline models, extensive evaluations across practically relevant tasks, and deliberate in-distribution and out-of-distribution splits to benchmark generalization. The results demonstrate strong performance for state-of-the-art MLIPs on many tasks, while highlighting persistent challenges in long-range interactions, charged/spin-state chemistry, and reactivity, guiding future architectural and sampling improvements. The work aims to catalyze community collaboration through public data, model weights, and planned leaderboards, accelerating the development of chemistry-wide ML surrogates for DFT-scale accuracy.

Abstract

Machine learning (ML) models hold the promise of transforming atomic simulations by delivering quantum chemical accuracy at a fraction of the computational cost. Realization of this potential would enable high-throughout, high-accuracy molecular screening campaigns to explore vast regions of chemical space and facilitate ab initio simulations at sizes and time scales that were previously inaccessible. However, a fundamental challenge to creating ML models that perform well across molecular chemistry is the lack of comprehensive data for training. Despite substantial efforts in data generation, no large-scale molecular dataset exists that combines broad chemical diversity with a high level of accuracy. To address this gap, Meta FAIR introduces Open Molecules 2025 (OMol25), a large-scale dataset composed of more than 100 million density functional theory (DFT) calculations at the $ω$B97M-V/def2-TZVPD level of theory, representing billions of CPU core-hours of compute. OMol25 uniquely blends elemental, chemical, and structural diversity including: 83 elements, a wide-range of intra- and intermolecular interactions, explicit solvation, variable charge/spin, conformers, and reactive structures. There are ~83M unique molecular systems in OMol25 covering small molecules, biomolecules, metal complexes, and electrolytes, including structures obtained from existing datasets. OMol25 also greatly expands on the size of systems typically included in DFT datasets, with systems of up to 350 atoms. In addition to the public release of the data, we provide baseline models and a comprehensive set of model evaluations to encourage community engagement in developing the next-generation ML models for molecular chemistry.

Figures (8)

• Figure 1: Overview of , including chemical scope, sampling strategies used to construct structures, chemical phenomena we seek to capture, properties available for each datapoint, and envisioned application areas.
• Figure 2: OMol25 dataset composition. a) Periodic table heat map, showing the number of snapshots in the training set containing a given element. b) Histogram in log scale for the number of training set snapshots with a given number of atoms. c) Histogram in log scale of training set snapshots with a given energy value relative to atomic references. d) Histogram in log scale of atomic force norms in the training set. e) Heat map for number of training set snapshots of different charge:spin.
• Figure 3: OMol25 breakdown by domain and sampling strategy by number of atoms. The percentage next to each domain indicates its share of the total dataset's number of atoms.
• Figure 4: evaluations. a) Ligand-pocket interaction energy/force, as defined by the energy/force difference between the ligand-pocket complex and the isolated ligand and isolated pocket. b) Ligand strain energy and conformer optimization/ordering, both of which involve a global optimization where many conformers of a structure are all subjected to a tightly converged geometry optimization. c) Relative protonation energies, as defined by the energy difference between optimized structures of different protonation states. d) Unoptimized ionization energy (IE) / electron affinity (EA) / spin gap, as defined by the energy difference between static structures of varying charge and/or spin multiplicity. e) Distance scaling, which seeks to predict the energy difference between instances of the same structure containing multiple molecular components, with the inter-component distance scaled by some factor.
• Figure 5: Histogram cross-sections of ligands sampled for metal complex datasets. a) total number of atoms in the ligands b) metal-coordinating atoms from the ligands c) denticity of the ligands and d) the total charges on the ligands.