The OpenFold Consortium has released a preview of OpenFold3, a fully open-source deep learning model that predicts the three-dimensional structures of proteins, nucleic acids, and their interactions with small molecule drugs and other biomolecules. Trained on over 300,000 experimentally determined structures and a synthetic database of more than 13 million structures, OpenFold3 represents a major leap in computational biology and serves as a foundational model for next-generation AI-driven tools in drug discovery, enzyme engineering, biosensor design, and biomaterials development. Unlike AlphaFold3, which is restricted to academic use and not licensed for commercial applications, OpenFold3 is released under the permissive Apache 2.0 license. This allows unrestricted use, modification, retraining, and integration into commercial pipelines—making it accessible to researchers, biotech firms, pharmaceutical companies, and academic institutions worldwide. OpenFold3 predicts protein folding from amino acid sequences alone and now extends its capabilities to model protein-ligand and protein-nucleic acid complexes—key interactions that underpin most modern drugs. This enables faster, more cost-effective in silico screening and design of therapeutics, significantly accelerating the R&D process. The model was developed by a global team led by the AlQuraishi Lab at Columbia University, the Bioresilience Program at Lawrence Livermore National Laboratory, and the Steinegger Lab at Seoul National University. It is built using PyTorch and optimized for high-performance inference through NVIDIA NIM, a containerized, accelerated API that supports scalable deployment across cloud and on-premise environments. A key design principle of OpenFold3 is modularity and usability. It can be easily adapted to work with proprietary data formats and integrated into diverse scientific workflows without requiring extensive technical overhead. This flexibility lowers the barrier to entry for data scientists and research teams across industries. Major companies are already adopting OpenFold3 for real-world applications. Novo Nordisk plans to integrate it into internal discovery pipelines using proprietary data. Outpace Bio will use it to design advanced cell therapies with engineered molecular circuits. Bayer Crop Science will apply it to study plant, weed, and pest proteins to develop new crop protection solutions. Cyrus Biotechnology aims to use the model to design enzyme-based therapies for autoimmune diseases. The open-source nature of OpenFold3 fosters collaboration and innovation. The U.K. government’s OpenBind initiative will fine-tune the model with new data, while the AI Structural Biology (AISB) Network is training it on datasets from major pharmaceutical companies. These efforts are part of a growing ecosystem supported by the Open Molecular Software Foundation (OMSF), which hosts OpenFold as a project. Industry partners including SandboxAQ, AWS, and others have contributed computational resources, funding, and expertise. AWS highlighted how its cloud infrastructure enabled rapid development and scalable training of the model at optimized cost, supporting a collaborative approach to advancing life sciences AI. OpenFold3 is available via GitHub, Hugging Face (with Docker images and model checkpoints), and hosted interfaces through partners Tamarind Bio and Apheris. The full list of contributors and member organizations is available at openfold.io. The consortium’s vision is to establish OpenFold as the standard in both academic and industrial research—mirroring the role of Linux in computing. As the community continues to build on this foundation, OpenFold3 is poised to become a cornerstone of the next wave of biological innovation.