AI for Bio · Model

RNAPro

Open model predicting all-atom RNA 3D structure, about 488M parameters

Structure & co-foldingMixedUnited StatesMaintained only中文

Overview

NameRNAPro · aliases: RNAPro
Full nameNo separate expanded name
Ambiguous names (not auto-linked)None
FamilyNot a family entry. Single-model entry, not a family
OrganizationNVIDIA, Stanford University
Organization typeMixed · Joint work by a big-tech lab and an academic lab
CountryUnited States
Model classRNA structure prediction
DomainStructure & co-folding (also: Genomics & RNA)
First release · code release
Basis: First commits on GitHub and Hugging Face on the same day source
Other milestones
  • Code release: · First GitHub commit source
  • Weights release: · First Hugging Face commit source
Latest update · maintenance only
Basis: README documentation update (post-processing guide, citation, data provenance)
Documentation only; no new weights or version source
The most recent sourced event, not a claim that nothing changed since.
StatusMaintained only · Documentation maintenance only since release
OpennessCode: Open · Apache-2.0
Weights: Open · NVIDIA Open Model License
Code Apache-2.0; weights under the NVIDIA Open Model License, commercial use allowed
Benchmarks reported1 counted (modified protocol 1)
Wet-lab validationNone recorded
Unverified fieldsNone
As of2026-11-16

What it does

Predicts all-atom RNA 3D structure by combining a co-folding architecture, RibonanzaNet2, MSA and template modeling at about 488M parameters; grew out of a public blind code competition entry.

Benchmarks reported

Every row is one pairing from the survey's relation file; a score belongs to the benchmark page's ranking, where it is shown with metric, eval set, source type and date. 1 of 1 pairings are counted.

BenchmarkRelationEval set / protocolSource typeNote and evidence
25 RNA-PuzzlesDeveloper report, modified protocolRNA-Puzzles subset within a competition private leaderboardSelf-reported10-target RNA-Puzzles subset inside Kaggle private leaderboard †partly verified · evidence

Other evaluations (outside the 44)

None recorded.

Developers report: Source states no developer claim

Independent evidence: No independent evidence found. No independent evidence found

Training data

Group in Figure 7: Public data only.

ClassValueDetail
a. Public databasesYesProtenix training set plus public RNA folding competition data (5,135 sequences with all-atom labels)
b. Private or licensed dataNo
c. Own wet-lab data used for trainingNo
d. Wet lab used only for validationNoEvaluation used a private set of newly solved structures

Wet-lab validation

None. The developers report no wet-lab validation.

Background

  • Organization: NVIDIA is a listed GPU and AI company; RNAPro belongs to its open model family and is internally funded; the academic side is a Stanford biochemistry team
  • Funding: Not disclosed or not applicable
  • Products and partnerships: Builds on Protenix and RibonanzaNet2; grew out of a blind competition with 1,700+ teams
  • Related models: Protenix, RhoFold+/DRfold2/trRosettaRNA

Reading notes

  1. Its only pairing is a subset inside a competition private leaderboard, not comparable with the official contest
  2. The weights licence allows commercial use, which is uncommon among comparable RNA structure models

Caveats

  1. Preprint posting date not verified
  2. The evaluation set is private newly solved structures and cannot be reproduced externally

Sources

  1. GitHub repository repos/NVIDIA-Digital-Bio (repository)
  2. Hugging Face page nvidia/RNAPro-Public-Best-500M (repository)
  3. GitHub repository NVIDIA-Digital-Bio/RNAPro (repository)
  4. bioRxiv preprint (preprint)

Corrections

None recorded.

Part of the AI for Bio benchmarks survey · all 112 models · open in the explorer. Record as of 2026-11-16.