ReleaseGoogle (DeepMind / Gemini)Google (DeepMind / Gemini)published Jun 9, 2026seen 1d

google-deepmind/alphafold3 v3.0.3

google-deepmind/alphafold3

Open original ↗

Captured source

source ↗
published Jun 9, 2026seen 1dcaptured 1dhttp 200method plain

AlphaFold v3.0.3

Repository: google-deepmind/alphafold3

Tag: v3.0.3

Published: 2026-06-09T12:26:50Z

Prerelease: no

Release notes: Version 3.0.3 changes the AlphaFold 3 code license from CC BY-NC-SA 4.0 to Apache 2.0. Moreover, it provides bug fixes and a performance improvement for large inputs. We thank everyone who reported issues, proposed new features, and submitted PRs!

Change log

Below is a selection of new features, performance improvements and bug fixes. Full change log: v3.0.2...v3.0.3

New features

  • AlphaFold 3 is now compatible with Python 3.14.
  • The AlphaFold 3 code is now relicensed from CC BY-NC-SA 4.0 to Apache 2.0.

Performance improvements

  • Confidences JSON serialization is up to 30x faster, for a 5,000-residue input the serialization time drops 43 -> 1.3 s.
  • mmCIF parsing is 4% faster.

Bug fixes

  • A new flag was introduced to not drop O1 atom from non-linking saccharides/glycans. Thanks @magnusbauer and @Jonas-Wilhelm.
  • Added support for dynamic version (inferred from the commit history) for dev and releases. Thanks @ccoulombe.
  • Pass all flags correctly to Jackhmmer/Nhmmer. Thanks @chenbq18.
  • Fix glycan O1 leaving-atom cleanup for standalone glycans. Thanks @Jonas-Wilhelm and @magnusbauer.
  • Do not drop chain descriptions for RNA and protein chains. Thanks @ntnn19.
  • Fixes of typing issues revealed by Pyrefly.

Documentation

  • Fixed typos in docstrings, thanks @mooreneural.

We thank the following people for their pull requests and/or great suggestions: @ccoulombe, @Jonas-Wilhelm, @magnusbauer, @mooreneural, @ntnn19.