Which molecular AI models run on Apple Silicon?
MoleculeDesk has verified local Boltz-2 installation and prediction on Apple Silicon using PyTorch MPS. The CLI detects your Mac's hardware with moldesk doctor. Model support is still in beta, and a model appearing in the registry does not mean its run path has been verified on every Mac.
Current MoleculeDesk status
| Model | Purpose | Apple Silicon status |
|---|---|---|
| Boltz-2 | Structure and binding-affinity prediction | Install and MPS prediction verified on Darwin arm64; beta |
| ProteinMPNN | Structure-conditioned sequence design | Beta adapter; check its model page and moldesk doctor for your environment |
| LigandMPNN | Ligand- and multimer-aware sequence design | Beta adapter; check its model page and moldesk doctor for your environment |
| OpenDDE Preview | All-atom biomolecular co-folding | Install and MPS prediction verified on Darwin arm64; beta (upstream preview) |
| DiffDock-L | Blind small-molecule docking | Coming soon; CUDA-focused path, not an Apple Silicon option in MoleculeDesk today |
AlphaFold, ColabFold, and Chai are listed as upcoming integrations; listing them is not a claim that they can currently be installed or run through MoleculeDesk. See model compatibility for status definitions.
Try the verified Boltz-2 path
curl -fsSL https://moleculedesk.com/install.sh | sh
moldesk doctor
moldesk install boltzContinue with the step-by-step Boltz-2 guide for a real input and run command. The Apple Silicon adapter uses the boltz-community MPS fork; Linux x64 uses upstream Boltz with its CUDA path. These are different pinned distributions, so hardware support and performance should not be inferred from one platform to the other.