AlphaFold 3
Predict the joint structure of proteins, nucleic acids and ligands
What's new
Model weights are available to academic/non-commercial groups by request; AlphaFold Server offers free browser-based prediction for non-commercial use.
AlphaFold 3 extends the AlphaFold line beyond single proteins to the joint structure of biomolecular complexes — proteins together with nucleic acids, ions, and small-molecule ligands. It replaces the Evoformer+structure-module design of AlphaFold 2 with a simplified Pairformer trunk and a diffusion module that generates atomic coordinates directly.
Highlights
- Models protein–protein, protein–nucleic acid and protein–ligand interactions in one network.
- Available for non-commercial use through the free AlphaFold Server web interface.
- Model parameters can be requested for academic use; code is open under Apache-2.0.
Good to know
The weights are restricted to non-commercial organisations and cannot be redistributed, which is why fully open alternatives such as Boltz, Protenix and OpenFold3 have become popular for commercial and reproducible workflows.
Related tools
AI Models
Boltz-2
Open-source structure and binding-affinity prediction, MIT licensed
AI Models
Protenix
High-accuracy open-source AlphaFold3 reproduction from ByteDance
AI Models
OpenFold3
A community, bitwise-faithful open reproduction of AlphaFold 3