CategoryKnowledge
MeasureComposite structural design score
Tasks15 protein targets
DifficultyLong-horizon computational protein design
Designs diverse miniprotein binders against target protein structures in a GPU-equipped research sandbox.
De novo protein-binder design ranking
2 models with a published De novo protein-binder design value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
Models ranked by their published Anthropic de novo protein-binder design evaluation value| Rank | Model | Provider | Composite structural design score |
|---|
| 1 |  Claude Opus 5.5 | Anthropic | 82.6 |
| 2 | Claude Sonnet 5.5 | Anthropic | 82.3 |
Evidence key: Observed
Rows are ordered by the value Claude Opus 5.5 System Card published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.
About De novo protein-binder design
What the benchmark measures and how its value is published, from the source’s own record.
What is De novo protein-binder design?
De novo protein-binder design (Anthropic de novo protein-binder design evaluation) is a knowledge benchmark published by Claude Opus 5.5 System Card. Designs diverse miniprotein binders against target protein structures in a GPU-equipped research sandbox. Its evaluation set is described by the source as 15 protein targets. The source rates its difficulty as long-horizon computational protein design.
How is De novo protein-binder design scored?
Each model's result is published by Claude Opus 5.5 System Card as composite structural design score. ModelScale prints that value to 1 decimal place, exactly as published, and never converts or rescales it. This page ranks the 2 models that carry a published value, highest value first; a model the source never scored is left out rather than ranked at zero. The source does not state whether a higher value is the better result, so neither does this page.
ModelScale