Protein Language Model Eval 2026
Consolidated evaluation harness for ESM, ProGen, xTrimoPGLM, Evo — 37 zero-shot tasks spanning fitness, structure, function.
Composite
100.0
Experimental validation
Wet-lab confirmed
Stages
Virtual CellHit ID
Modalities
biologic
Task types
zero-shot-fitnessrepresentation-quality
Size
tasks: 37
proteins: 9,800,000
proteins: 9,800,000
License
MIT
First release
2026-02
Last updated
2026-04
Official site
Leaderboard
Dataset
→ dataset
Code / GitHub
HuggingFace
Paper
Protein LM Eval 2026: a community harness · Rives A, Rao R, et al. · 2026 · paper · doi:10.1101/2026.02.10.550123 · 76 citations
Flags
none
Experts
—
Groups
—
Hosted by
—
Related benchmarks
Rubric (7-criterion)
rigor
5
coverage
5
maintenance
5
adoption
5
quality
5
accessibility
5
industry_relevance
5
Notes
Meta FAIR + EvolutionaryScale collaboration; includes held-out targets with wet-lab fitness.