Skip to main content

​
Obtaining Model Parameters

To use AlphaFold 3 for structure prediction, you need to obtain the model parameters (weights) from Google DeepMind.
Important: You may only use AlphaFold 3 model parameters if received directly from Google. Use is subject to the AlphaFold 3 Model Parameters Terms of Use.

​
Request Access

1

Complete the access request form

Visit the Google Form to request access to AlphaFold 3 model parameters:Request AlphaFold 3 Model ParametersYou will need to provide:
  • Your name and institutional affiliation
  • Email address
  • Intended use case
  • Agreement to the terms of use
2

Wait for approval

Access will be granted at Google DeepMind’s sole discretion.
Expected Response Time: 2-3 business days
You will receive an email notification once your request has been processed.
3

Download the model parameters

Once access is granted, you will receive instructions for downloading the model parameters.Download them to a directory on your system (referred to as <MODEL_PARAMETERS_DIR> in the documentation).
4

Store model parameters

Store the model parameters in a location that is:
  • NOT a subdirectory of the AlphaFold 3 repository
  • Accessible to your Docker container
  • Has sufficient disk space
Choose a location like $HOME/alphafold3_models or /data/alphafold3/models that will be easy to mount in Docker.

​
Terms of Use

Before requesting access, carefully review the AlphaFold 3 Model Parameters Terms of Use.

​
Key Terms

  • Non-Commercial Use: The model parameters are subject to specific usage restrictions
  • No Redistribution: You may not share or redistribute the model parameters
  • Direct from Google: Parameters must be obtained directly from Google DeepMind
  • Citation Required: Any publication using the model parameters must cite the AlphaFold 3 paper
  • No Clinical Use: Not intended, validated, or approved for clinical applications

​
Prohibited Uses

The model parameters have specific prohibited use cases. Review the Prohibited Use Policy for complete details.

​
Using Model Parameters

Once you have the model parameters, you can use them with AlphaFold 3:

​
With Docker

docker run -it \
    --volume $HOME/af_input:/root/af_input \
    --volume $HOME/af_output:/root/af_output \
    --volume <MODEL_PARAMETERS_DIR>:/root/models \
    --volume <DB_DIR>:/root/public_databases \
    --gpus all \
    alphafold3 \
    python run_alphafold.py \
    --json_path=/root/af_input/fold_input.json \
    --model_dir=/root/models \
    --output_dir=/root/af_output
The --model_dir flag should point to the directory containing your model parameters (mounted at /root/models inside the container).

​
With Singularity

singularity exec \
     --nv \
     --bind $HOME/af_input:/root/af_input \
     --bind $HOME/af_output:/root/af_output \
     --bind <MODEL_PARAMETERS_DIR>:/root/models \
     --bind <DB_DIR>:/root/public_databases \
     alphafold3.sif \
     python run_alphafold.py \
     --json_path=/root/af_input/fold_input.json \
     --model_dir=/root/models \
     --db_dir=/root/public_databases \
     --output_dir=/root/af_output

​
Alternative: AlphaFold Server

If you don’t need the full flexibility of the local installation, you can use AlphaFold 3 through the web interface:

AlphaFold Server

Use AlphaFold 3 online for non-commercial research without requiring model parameters or local installation.Available at: alphafoldserver.comNote: The server has a more limited set of ligands and covalent modifications compared to the local installation.

​
When to Use AlphaFold Server

Use the Server When:

  • You need quick predictions for standard cases
  • You don’t have access to GPU infrastructure
  • Your use case is covered by available ligands
  • You prefer a web interface

Use Local Installation When:

  • You need custom ligands or SMILES
  • You require high throughput predictions
  • You need full control over MSA and templates
  • You want to run on your own infrastructure
  • You need complex covalent modifications

​
Citation Requirements

Any publication that discloses findings arising from using the model parameters or outputs produced by them must cite the AlphaFold 3 paper.

​
BibTeX Citation

@article{Abramson2024,
  author  = {Abramson, Josh and Adler, Jonas and Dunger, Jack and Evans, Richard and Green, Tim and Pritzel, Alexander and Ronneberger, Olaf and Willmore, Lindsay and Ballard, Andrew J. and Bambrick, Joshua and Bodenstein, Sebastian W. and Evans, David A. and Hung, Chia-Chun and O'Neill, Michael and Reiman, David and Tunyasuvunakool, Kathryn and Wu, Zachary and Žemgulytė, Akvilė and Arvaniti, Eirini and Beattie, Charles and Bertolli, Ottavia and Bridgland, Alex and Cherepanov, Alexey and Congreve, Miles and Cowen-Rivers, Alexander I. and Cowie, Andrew and Figurnov, Michael and Fuchs, Fabian B. and Gladman, Hannah and Jain, Rishub and Khan, Yousuf A. and Low, Caroline M. R. and Perlin, Kuba and Potapenko, Anna and Savy, Pascal and Singh, Sukhdeep and Stecula, Adrian and Thillaisundaram, Ashok and Tong, Catherine and Yakneen, Sergei and Zhong, Ellen D. and Zielinski, Michal and Žídek, Augustin and Bapst, Victor and Kohli, Pushmeet and Jaderberg, Max and Hassabis, Demis and Jumper, John M.},
  journal = {Nature},
  title   = {Accurate structure prediction of biomolecular interactions with AlphaFold 3},
  year    = {2024},
  volume  = {630},
  number  = {8016},
  pages   = {493--500},
  doi     = {10.1038/s41586-024-07487-w}
}

​
Plain Text Citation

Abramson, J., Adler, J., Dunger, J. et al. Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 630, 493–500 (2024). https://doi.org/10.1038/s41586-024-07487-w

​
Frequently Asked Questions

Google DeepMind aims to respond to requests within 2-3 business days. However, processing times may vary depending on request volume.
No. The model parameters must be obtained directly from Google and may not be redistributed. Each user must request their own access.
Access is granted at Google DeepMind’s sole discretion. If your request is denied, you can:
The model parameters are subject to specific terms of use. Review the Terms of Use and Prohibited Use Policy for details on permitted uses.
The model parameters download size and disk space requirements will be specified when you receive access. Ensure you have sufficient storage before downloading.
If new versions of the model parameters are released, check the instructions provided with your original access grant for information about obtaining updates.

​
Troubleshooting

​
Model Directory Not Found

If you get an error about the model directory: 
# Verify the directory exists and contains model files
ls -la <MODEL_PARAMETERS_DIR>

# Check Docker volume mount
docker run -it \
    --volume <MODEL_PARAMETERS_DIR>:/root/models \
    alphafold3 \
    ls -la /root/models

​
Permission Issues

# Ensure the model directory has appropriate permissions
chmod 755 <MODEL_PARAMETERS_DIR>

​
Next Steps

Once you have obtained the model parameters:

Complete Installation

Finish setting up AlphaFold 3 with databases and Docker

Run Your First Prediction

Start making structure predictions

​
Support

For questions about model parameter access:
Please do not create GitHub issues about access requests. Contact [email protected] directly for questions about the access process.

Build docs developers (and LLMs) love

Get started for freeTalk to us