Apple Silicon · MLX · Metal

Atomistic simulation,
native to your Mac.

An alpha MD and DFT runtime built directly on MLX and Metal — it runs the GPU on your machine, not a wrapper around someone else's cluster.

Get started View on GitHub

Capabilities

One runtime, two scales.

Apple Silicon native

MLX arrays execute through the Metal backend on your machine. No CUDA, no server, no cloud.

Plane-wave DFT

Γ-point Kohn–Sham SCF, LDA plus public-alpha PBE-PZ81 GGA diagnostics, non-SCF band paths.

Molecular mechanics

Lennard-Jones, Coulomb, harmonic bonds and angles, periodic + RB torsions, bounded PME, Langevin NVT.

Pseudopotentials

GTH and UPF readers, proof-level projector diagnostics, analytic local forces plus fixed-orbital nonlocal diagnostics.

Python 3.13 · uv

Reproducible environments, fast sync, Jupyter-first exploration.

Reference validation lanes

OpenMM and LAMMPS remain opt-in local reference surfaces, not package runtime dependencies.

LDA · PBE-PZ81

XC diagnostics

NVE · NVT

Integrators

PME

Long-range electrostatics

Jupyter

First-class visualization

Quick start

Try the alpha, then open a checkout.

zsh — mlx-atomistic

# try the PyPI alpha
$ uv run --no-project --python 3.13 --with mlx-atomistic \
    python -c "import mlx_atomistic as ma; print(ma.__version__)"

# work from a checkout
$ uv venv --python 3.13
$ uv sync --extra notebook --extra prep --extra viz
$ uv run jupyter lab

# run a DFT SCF benchmark
$ uv run python -m mlx_atomistic.benchmarks.dft_scf \
    --sizes 8,16,24,32 --iterations 5 --mixer both --json

Atomistic simulation, native to your Mac.