Google DeepMind Fully Open-Sources AlphaFold 3 Code and Weights
DeepMind officially open-sources AlphaFold 3’s core code and model weights, providing free access for non-commercial academic research.
Open-Source Decision Becomes Official
According to the journal Nature, Google DeepMind has made the core code and trained model weights of AlphaFold 3 fully open-source. Previously, AlphaFold 3 was only available with limited functionality through an online server, with strict restrictions on usage frequency and commercial applications. With this release, researchers worldwide can freely download, use, and modify the code and weights for non-commercial academic projects.
AlphaFold 3’s Technological Breakthroughs
AlphaFold 3 is the next-generation protein structure prediction model from DeepMind, following AlphaFold 2. It not only predicts protein structures but also handles interactions with other molecules such as DNA, RNA, small-molecule drugs, and ions, significantly expanding its predictive scope. The model shows a marked improvement in accuracy over its predecessor, demonstrating powerful capabilities, especially in modeling multi-molecular complexes.
Impact on Scientific Research
The open-source initiative is seen as a major step forward for structural biology and drug development. Researchers can now deploy the model locally, bypassing server queues and fine-tuning or extending it for specific research needs. This will help accelerate work in various fields, including new drug discovery, enzyme engineering design, and fundamental biological mechanism research.
Access and Usage Restrictions
Although the code and weights are fully open-source, DeepMind specifies that their use is restricted to non-commercial academic purposes. Any commercial application still requires a license from DeepMind or Google’s official channels. This move aims to balance scientific openness with the protection of intellectual property.
Continuing the Open-Source Tradition
DeepMind has consistently promoted scientific openness since the inception of the AlphaFold project. The open-sourcing of AlphaFold 2 greatly facilitated thousands of biomedical studies worldwide, and this release of AlphaFold 3 is seen as a continuation of that open strategy. In the future, more derivative tools and improved versions based on AlphaFold 3 are expected to emerge rapidly within the open-source community.
