AWS adds the OpenFold Consortium’s all-atom model to its managed catalog, targeting drug discovery teams that lack GPU plumbing.
AWS added OpenFold3-preview to Amazon SageMaker JumpStart, giving teams a managed way to deploy the open source all-atom structure prediction model. Developers submit protein, DNA, RNA, and ligand inputs and receive predicted 3D structures in mmCIF format along with confidence metrics.
The OpenFold Consortium and the AlQuraishi Lab at Columbia University developed OpenFold3 and release its code, weights, and training methods under the Apache 2.0 license. The version on JumpStart is OpenFold3 0.4.1, which runs the of3-p2-155k checkpoint, and the developers label it a research preview.
Operators get GPU infrastructure, a CUDA and PyTorch stack, and request queuing that tolerates jobs running in minutes rather than milliseconds. Network isolation keeps sequences inside the AWS account, and teams can supply their own multiple sequence alignments to raise accuracy or scale the endpoint to zero when idle.
Structure prediction teams use these models for target identification, antibody and protein engineering, and structure-based virtual screening. The research preview runs on ml.g6 instances, and multi-GPU options let teams run several predictions in parallel while they watch for full production support.
What matters
- AWS now hosts OpenFold3-preview on SageMaker JumpStart, letting teams deploy it with a few lines of code.
- The managed endpoint handles GPU warmup, request queuing, and autoscaling, removing setup work for operators.
- Watch whether the research preview extends ligand coverage beyond the current of3-p2-155k checkpoint.
Why it matters
Watch whether the research preview extends ligand coverage beyond the current of3-p2-155k checkpoint.
This GenAI News article was prepared in original wording using reporting and materials published by AWS Machine Learning Blog. Source reference: https://aws.amazon.com/blogs/machine-learning/openfold3-on-amazon-sagemaker-jumpstart-predict-biomolecular-structures/.
Drafted by the GenAI News review pipeline.