← Datasets & benchmarks

FrontierBioML

FrontierBioML is a datasets & benchmarks tool, first public in 2026. Valen Labs benchmark of 25 long-horizon ML tasks in computational biology (antibody development, protein engineering, structural biology, binder design, functional genomics): an agent works offline for up to ~6 hours, submits a prediction script, and is scored on held-out experimental data.

Visit website ↗

CategoryDatasets & benchmarks
Websitewww.frontierbioml.com
Released2026-11-12 (source)
LicensePublic preview of 4 of 25 tasks on Harbor and Prime Intellect Environment Hub (full set on request); each task redistributes third-party data under its own licence (t01 AVIDa-hIL6 CC BY-NC 4.0 non-commercial; t09, t22, t25 CC BY 4.0); held-out labels and scorers are not released
What this allowsUnknown
No confirmed terms from a primary source; ask the authors before using it in a product. What do these mean?
License sourcehub.harborframework.com/datasets/valenlabs/frontierbioml · checked 2026-11-14
PricingResearch release
Tags#benchmark#agents#antibody#protein-design#functional-genomics
Last checked · links verified

Can I use FrontierBioML commercially?

Unclear. We could not confirm the terms from a primary source; check with the authors before commercial use.

What the license types mean · not legal advice.

Specs

Size25 tasks (4 public preview tasks: t01, t09, t22, t25)
Modality—
AccessLeaderboard at frontierbioml.com; preview tasks via harbor run -d valenlabs/frontierbioml and Prime Intellect Environment Hub
Interface—

Compare with…

Pick 1–3 other Datasets & benchmarks tools to compare side by side with FrontierBioML.

Select 1–3 tools

Related

OMol25

Meta FAIR's open dataset of DFT calculations on diverse molecules, with evaluations and baseline models.

Datasets & benchmarksOpen sourceUnknown#dft

ML-ready datasets and benchmarks for therapeutics.

Datasets & benchmarksOpen sourcePermissive (commercial OK)#benchmark

ProteinGym

Benchmarks for protein fitness and mutation effect prediction.

Datasets & benchmarksOpen sourcePermissive (commercial OK)#benchmark