jurgjn/alphafold3 repository overview

⁠Vanilla builds⁠, tags match commit/version from google-deepmind/alphafold3⁠

• v3.0.4⁠ from Jul 28 2026
• v3.0.3⁠ from Jun 9 2026
• v3.0.2⁠ from Apr 20 2026
• 608edb6⁠ from Mar 10 2026
• ebfe70a⁠ from Feb 4 2026
• b78e215⁠ from Jan 9 2026 (supports Blackwell GPUs⁠, e.g. RTX PRO 6000)
• a8ecdb2⁠ from Sep 8 2025
• 2e2ffc1⁠ from Aug 7 2025
• 0422023⁠ from Jul 14 2025
• v3.0.1⁠ from Jan 23 2025 (latest release⁠, default in batch-infer⁠)
• v3.0.0⁠ from Nov 11 2024

⁠Custom builds

• 97d66d2⁠ from sokrypton/alphafold3⁠ to use OpenFold3 weights⁠
• v3.0.1-daint1⁠ for arm64/gh200 (Grace Hopper) by modifying a single line in pyproject.toml from jurgjn/alphafold3⁠ (and removing all hashes from dev-requirements.txt)

⁠See also

• Singularity Containers for Protein Prediction Models⁠
• Getting AlphaFold 3 to run on AMD GPUs⁠
• Source Code Update for GPU Blackwell and CC 12.0 Compatibility in AlphaFold 3⁠