Model purpose

RFdiffusion is described as a generative model of protein backbones for protein binder design. Rather than predicting sequences, it addresses the structural stage of design by generating backbone geometries that a binder can be built around. Searches for rfdiffusion protein backbone generation map directly to this function within the IPD listing.

  • Generative model of protein backbones
  • Applied to protein binder design
  • Listed under the IPD model collection

Source: AI Models by IPD | Try NVIDIA NIM APIs

Listing and classification

On the IPD page, rfdiffusion appears as a downloadable model. Its classification places it in the biology domain, with further tags covering NIM packaging, BioNeMo, protein generation and drug discovery. These tags describe how the model is grouped alongside other biology models in the catalogue.

Availability
downloadable
Domain tag
biology
Additional tags
nim, BioNemo, Protein Generation, Drug Discovery

Source: AI Models by IPD | Try NVIDIA NIM APIs

Related models from IPD

The IPD collection lists two models in total. Alongside rfdiffusion sits ProteinMPNN, a deep learning model for predicting amino acid sequences for protein backbones. The pairing covers backbone generation and sequence prediction for those backbones.

  • rfdiffusion — protein backbone generation
  • proteinmpnn — amino acid sequence prediction for protein backbones

Source: AI Models by IPD | Try NVIDIA NIM APIs

Access

The model is reached through the IPD publisher page on NVIDIA Build, where the available models are listed together with their tags and download status. The same page presents the NIM API route used to try the listed models.

  • Publisher page: build.nvidia.com/ipd
  • Offered via NVIDIA NIM APIs

Source: AI Models by IPD | Try NVIDIA NIM APIs

Sources

  1. AI Models by IPD | Try NVIDIA NIM APIs https://build.nvidia.com/ipd Verified 03 Nov 2026

Last verified 02 Oct 2026. This entry is compiled from the public web pages listed above. Nothing here is stated that those pages do not, and each of them was read on the date shown.