The Open Materials 2024 (OMat24) inorganic materials dataset and models.

Open Materials 2024 dataset enables machine learning models to surpass previous standards, achieving F1 scores above 0.9 and 20 meV per atom accuracy in inorganic materials.

• Why it matters: Advancing the discovery and simulation of inorganic materials is crucial for addressing global challenges like climate change and semiconductor development, yet accessible datasets lag behind proprietary models.
• What they did: The team compiled over 110 million density functional theory calculations into the OMat24 dataset and trained interatomic potential models, testing their performance on established benchmarks.
• The result: Models trained on OMat24 demonstrate high accuracy in property predictions, correcting prior biases and enabling significant improvements in inorganic materials research and applications.