What changed
AWS has announced the availability of OpenFold3 through Amazon SageMaker JumpStart, allowing users to deploy the model with just a few lines of code. OpenFold3 is an all-atom structure prediction model developed by the OpenFold Consortium and the AlQuraishi Lab at Columbia University. It can predict the 3D structures of proteins, DNA, RNA, small-molecule ligands, and the complexes they form. The model runs in an AWS-maintained container on a managed Amazon SageMaker AI endpoint, which handles request queuing, GPU warmup, and autoscaling.
What this means in practice
Users can deploy OpenFold3-preview on an asynchronous inference endpoint, submit structure prediction jobs, and interpret results including predicted structures and confidence scores. The model supports multi-GPU instances for parallel predictions and allows users to supply their own multiple sequence alignments (MSAs) to improve accuracy while keeping sequences private. The endpoint can scale to zero instances when idle to reduce costs. Computational drug discovery teams can use this for target identification, antibody engineering, and structure-based virtual screening.
Limitations and availability
OpenFold3 runs on supported instances ml.g6.2xlarge, ml.g6.4xlarge (default), and ml.g6.12xlarge. It is available under the Apache 2.0 license, with the model ID huggingface-proteinfolding-openfold3. The developers describe it as a research preview, and it is designed for asynchronous inference due to the long processing times required for structure prediction. Users must have an AWS account, a SageMaker AI execution role, and access to SageMaker Studio or a local IDE. The model is available in regions with a service quota for at least one ml.g6.4xlarge endpoint instance.
