OpenFold3

5-30 minutesPaperCodeNeed >20GB

AlphaFold3 reproduction, predict protein/nucleic acid/small molecule complexes

Readiness

Readiness: Need >20GB
Smoke: passed
Artifacts: passed
Production: need_gt20gb_openfold3_msa_databases_or_precomputed_scope
Storage: >20GB
Strategy: OpenFold3 SIF/pixi runtime plus checkpoint and explicit MSA policy; full offline MSA database setup is hundreds of GB to TB.
Next: Provision database storage for OpenFold3 MSA databases or approve a server/precomputed-MSA production scope; then stage SIF/checkpoint and run tiny GPU Slurm/API validation.

large model weights, structure DBs, docking/MD assets, or GPU stack likely >5GB 2026-07-16T20:35:00+08:00 SOP Need >5GB next-10 audit: remains Need >5GB. Registry maps to openfold3-smoke and runner uses generic shared smoke path, not an OpenFold3 SIF/model/database stack. 2026-07-18T19:02:14+08:00 SOP Need >5GB active install batch: Official OpenFold3 setup is public, but full offline production MSA generation needs default protein MSA databases around 330GB, with optional BFD/ColabFold/RNA databases much larger. Reclassified from Need >5GB to Need >20GB unless a narrowed precomputed/server-MSA scope is approved.