Openfold2
Openfold2
Version: 3
Nvidia•Last updated May 2025
Openfold2 is a protein structure prediction model from the OpenFold Consortium and the Alquraishi Laboratory . The model is a PyTorch re-implementation of Google Deepmind’s AlphaFold2 , with support for both training and inference. OpenFold2 demonstrates parity accuracy with AlphaFold2, and improved speed, see the project home for more detail aqlaboratory/openfold . The NVIDIA OpenFold2 NIM can:
  • Predict a protein structure given an input protein sequence, and accepts optional inputs such as multiple sequence alignments and templates.
This NIM implements the ‘monomer’ version of OpenFold2, and uses the model parameter sets trained with Google Deepmind’s original jax implemenation of AlphaFold2:
  • params_model_1.npz
  • params_model_2.npz
  • params_model_3.npz
  • params_model_4.npz
  • params_model_5.npz
For more information about OpenFold2, see the OpenFold2 paper in Nature .

Intended Use

Primary Use Cases

NIMs offer a simple and easy-to-deploy route for self-hosted AI applications. Two major advantages that NIMs offer for system administrators and developers are:
  • Increased productivity: NIMs allow developers to build generative AI applications quickly, in minutes rather than weeks, by providing a standardized way to add AI capabilities to their applications.
  • Simplified deployment: NIMs provide containers that can be easily deployed on various platforms, including clouds, data centers, or workstations, making it convenient for developers to test and deploy their applications.
The OpenFold2 NIM provides a fast, accurate model behind a consistent API for predicting protein structure. As part of the broader NVIDIA NIM Ecosystem, OpenFold2 can be used in conjunction with other NIMs to build pipelines that generate and assess the structure and function of entirely new proteins and small molecules.

Responsible AI Considerations

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their supporting model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse. Please report security vulnerabilities or NVIDIA AI Concerns here .

Training Data

The model parameter sets were trained by Google Deepmind as part of AlphaFold2 development. A description of the training dataset and relevant download links are available at Highly Accurate ... Data Availability . This data was not collected by NVIDIA. Data Collection Method by dataset Labeling Method by dataset Properties (Quantity, Dataset Descriptions, Sensor(s)): Uniclust dataset of 355,993 sequences with the full MSAs. These predictions were then used to train a final model with identical hyperparameters, except for sampling examples 75% of the time from the Uniclust prediction set, with sub-sampled MSAs, and 25% of the time from the clustered PDB set.
Openfold2 NIM is optimized to run best on the following compute:
GPUTotal GPU memoryAzure VM compute#GPUs on VMLink
A10080Standard_NC24ads_A100_v41link
A100160Standard_NC48ads_A100_v42link
A100320Standard_NC96ads_A100_v44link
A100640STANDARD_ND96AMSR_A100_V48link
H10094STANDARD_NC40ADS_H100_V51link
H100188STANDARD_NC80ADIS_H100_V52link
Model Specifications
LicenseCustom
Last UpdatedMay 2025
PublisherNvidia
Quick Start
Sign in to get startedView DocumentationAzure Support