OpenFold Consortium Releases OpenFold3 Update with Public Training Data for Reproducible Biomolecular AI

on

The OpenFold Consortium announced a major OpenFold3 update on March 13, 2026, releasing training datasets via AWS Registry of Open Data, along with model weights, training/inference code, and evaluation scripts under permissive licenses1.

OpenFold3 predicts 3D structures of biomolecular complexes, including proteins with small molecules and nucleic acids, showing competitive performance against AlphaFold3 in benchmarks1.

This full-stack release enables independent reproduction, benchmarking, retraining, and extension for drug discovery and research1.

New OpenFold3 portal provides onboarding, documentation, and community support channel1.

SandboxAQ integrated AQAffinity for rapid binding affinity prediction, enhancing OpenFold3 for early-stage drug discovery2.

OpenFold3 builds on a October 2025 preview, with Apache 2.0 licensing allowing broad use unlike restricted AlphaFold33.

Sources:

1. https://www.businesswire.com/news/home/20260313170622/en/OpenFold-Consortium-Announces-Major-OpenFold3-Update-and-Public-Release-of-Training-Data-for-Reproducible-Biomolecular-AI

2. https://www.prnewswire.com/news-releases/sandboxaq-supercharges-openfold3-with-aqaffinity-high-speed-structure-free-drug-potency-prediction-302665496.html

3. https://www.businesswire.com/news/home/20251028507233/en/OpenFold-Consortium-Releases-Preview-of-OpenFold3-An-Open-Source-Foundation-Model-for-Structure-Prediction-of-Proteins-Nucleic-Acids-and-Drugs