Therapeutic Antibody Design Benchmark 2026

Consolidated antibody-design benchmark covering developability, CDR diversity, paratope prediction, and experimental binding on 12 targets.

Composite

97.0

Experimental validation

Wet-lab confirmed

Stages

Hit IDLead ID / ADMET

Modalities

biologic

Task types

binder-designdevelopabilityaffinity-prediction

Size

antibodies: 85,000
targets: 12

License

CC-BY-4.0

First release

2026-02

Last updated

2026-04

Official site

→ project page

Leaderboard

→ leaderboard

Dataset

→ dataset

Code / GitHub

→ repository

HuggingFace

→ HF

Paper

Benchmarking machine-learning antibody design 2026 · Ruffolo JA, Chung LK, et al. · 2026 · paper · doi:10.1038/s41586-026-09876-1 · 65 citations

Flags

none

Experts

—

Groups

Genentech gRED (ML)

Hosted by

Polaris Hub

Related benchmarks

SAbDab, Observed Antibody Space (OAS), IgLM / AntiBERTa benchmarks

Rubric (7-criterion)

rigor

coverage

maintenance

adoption

quality

accessibility

industry_relevance

Notes

Top-ranked submissions had wet-lab binding measured (Kd + aggregation) by independent labs.

← Back to all benchmarks

Compare:

Open comparison →