MODEL INTELLIGENCE / VERIFIED RESEARCH
RFdiffusion3NA (RFD3NA)
Extension of RFD3 for jointly designing protein, DNA and RNA structures.
Developer: Institute for Protein Design / RosettaCommons, University of Washington
Overview
- Developer
- Institute for Protein Design / RosettaCommons, University of Washington
- Category
- de novo binder and protein design
- Commercial rights
- Commercial OK
- License fee
- No license fee identified
- Open weights
- Yes
- Last verified
- 2026-10-16
- Added to FoldScout
- 2026-10-16
Capabilities
Foundry's RFD3NA supports generative design of multi-polymer structures comprising proteins and nucleic acids with atomic constraints. It releases training/inference code and referenced checkpoints via Foundry. Optional hydrogen-bond conditioning may require external HBPLUS.
When to use it
Explore nucleic acid and protein co-design and mixed interfaces, not standard single-chain antibody sequence ranking. Optional external tools require separate license validation.
Compute
Compute classification: single-gpu. This is a planning estimate, not a throughput benchmark; check the linked model installation documents for exact requirements.
Commercial licensing
Commercial permission — Commercial OK
The Foundry BSD-3-Clause license includes code and the referenced checkpoints. Commercial use of the core model is permitted; optional HBPLUS and other external tools must be checked independently before their use.
License fee — No license fee identified
No Foundry model license fee. HBPLUS, computational resources and additional dependencies may have separate terms/costs; no blanket free claim is made.
Free downloads do not necessarily permit company use. Code, checkpoints, external dependencies and compute expenses require separate review.
Official references
From computational design to experimental evidence
Validate your designed binder.
Seq2Study connects AI-designed proteins and antibodies with expression, affinity, stability and developability testing.