Benchling Model Hub
Run protein and complex structure prediction (AlphaFold 2, Chai-1, Boltz-2, Protenix, OpenFold, BoltzGen) inside the Benchling notebook and registry.
Overview
Benchling Model Hub runs structure-prediction and generative models directly inside Benchling's electronic lab notebook and registry, where many biotech teams already keep their sequence and construct data. It is built for running inference in the flow of experimental work rather than distributing weights: models operate on registry inputs, execute on managed GPU, and record a full audit trail. The differentiator is integration and provenance inside the ELN, not a standalone model endpoint.
What you can run on Benchling Model Hub
The hub focuses on protein and biomolecular complex structure prediction. AlphaFold 2, OpenFold, and Chai-1 predict protein and complex structures; Protenix and Boltz-2 extend to co-folding and biomolecular complex prediction; and BoltzGen (in beta) adds generative structure modeling. All run as in-app inference on structures and sequences pulled straight from the registry, covering the core structure-prediction workload biologists reach for at the bench without leaving the notebook.
Running models inside the Benchling notebook
Models run in-platform: you select registry entities as inputs and launch single or batch jobs on managed compute, with results written back into Benchling and versioned alongside your data. There are no public per-model pages, weight downloads, or fine-tuning here; the value is turnkey inference with governance built into the notebook. A sibling product, Benchling Inference (powered by Baseten), provides GPU capacity to run and train models for teams that need to go beyond the hosted hub.
Run inference on Benchling Model Hub (6)
Protein structure prediction model that folds amino acid sequences into 3D structures with atomic accuracy, scoring a median GDT of 92.4 at CASP14.
Open model that jointly predicts biomolecular structure and small-molecule binding affinity, approaching FEP+ accuracy in seconds on a single GPU.
Trainable, open-source reimplementation of AlphaFold2 for protein structure prediction that matches its accuracy and runs 3-5x faster.
Biomolecular structure prediction foundation model covering proteins, small molecules, DNA, RNA, and glycans in a single diffusion framework.
Open-source PyTorch reproduction of AlphaFold 3 under Apache 2.0, matching or exceeding AF3 on protein-ligand, protein-protein, and RNA benchmarks.
All-atom generative model for de novo protein and peptide binder design against diverse biomolecular targets, wet-lab validated across 26 targets.