MatterSim
MicrosoftMatterSim is Microsoft's deep-learning atomistic model for bulk materials, acting as a machine-learning force field across the periodic table for temperatures from 0 to 5000 K and pressures up to 1000 GPa. It ships 1M- and 5M-parameter M3GNet checkpoints in the repository, with a fine-tuning script for adapting the model to a custom dataset or a higher level of theory.
The pretrained checkpoints ship inside the repository and through the mattersim PyPI package; more advanced versions are held back behind Azure Quantum Elements. Verified 2026-09-10 via the GitHub API, the repository README, and the model card.
Openness
3 high confidence- weights
- open(MIT MatterSim-v1.0.0-1M and MatterSim-v1.0.0-5M checkpoints in the repository)
- data
- closed(the 3M and 6M-structure training sets are not released
- code
- partial(inference and fine-tuning scripts
- license
- MIT(OSI)
The released v1 checkpoints are MIT and locally runnable, but the training data stays inside Microsoft and the repository documents fine-tuning rather than a pretraining run.
- https://raw.githubusercontent.com/microsoft/mattersim/main/MODEL_CARD.md recorded 2026-09-10
license: mit for the released models, training data sizes of 3M and 6M structures with no download, and released test datasets only.
- https://raw.githubusercontent.com/microsoft/mattersim/main/README.md recorded 2026-09-10
Offers two pretrained MatterSim-v1 checkpoints from the repository's pretrained_models folder and a finetune_mattersim.py script, with no pretraining entry point.
Adoption
1 medium confidenceNo Hugging Face artifact is declared, so 594 GitHub stars are the last-resort signal; the mattersim PyPI package is the channel that actually ships the weights and is not banded here.
- https://api.github.com/repos/microsoft/mattersim recorded 2026-09-10
stargazers_count: 594; archived: false; fork: false.
Capability
2 high confidenceOne attribute. The model card rules out generation explicitly, the temperature and pressure ranges are conditions rather than scales, and the 1M and 5M checkpoints are sizes.
- https://raw.githubusercontent.com/microsoft/mattersim/main/MODEL_CARD.md recorded 2026-09-10
States the model can be fine-tuned for atomistic simulations at a desired level of theory or for direct structure-to-property predictions, and that it does not support generation of new materials.
Verified 2026-09-10