AIEvals

Benchmarks

Protocols (understanding V2)

Answering practical questions about applying or adapting real laboratory protocols in Benchling’s Understanding V2 task set.

Life sciencesReported by AnthropicVendor-reported onlyReported by Anthropic

How it is run

Section 8.13.6 uses protocols.io documents, vendor manuals or facility guides, generally averaging five max-effort attempts with Haiku biology safeguards off.

What it cannot tell you

Version 2 understanding is separate from troubleshooting and earlier task versions. Scores do not measure real-world experimental success.

Why there is no independent table

No independent suite in this dataset runs Protocols (understanding V2). The numbers below are the ones model publishers report about their own models, which is a different kind of evidence: each publisher chooses its own harness and effort level, so a row is readable on its own and not against the other rows.

Read this with

Other versions of the same evaluation. Versions are not comparable: the task set changes between them.

  • Protocols (troubleshooting)Same family. Detecting and fixing problems in molecular-biology laboratory protocols under a publisher-defined troubleshooting evaluation.
  • SpatialBench VerifiedSame category. Analyzing spatial transcriptomics and chromatin-accessibility data in challenging real-world bioinformatics tasks from LatchBio.
  • SingleCellBenchSame category. Analyzing single-cell RNA sequencing data in real-world bioinformatics workflows developed by LatchBio.
  • Morphology-to-molecule matchingSame category. Matching four sets of multi-channel liver-cell images to their shuffled molecule-dose perturbations in Axiom Bio’s assay.
  • Medicinal chemistry (ADME)Same category. Selecting compounds with the highest or lowest measured absorption, distribution, metabolism or excretion properties.
  • Protein design (sequence generation)Same category. Generating novel protein sequences that satisfy a variety of design constraints in a protein-design evaluation.

As reported by each publisher

Kept separate per publisher, because the settings are not shared.

Anthropic

Claude Haiku 5.5: system card capability evaluations · 7 October 2026

RankModelScore
2Claude Haiku 5.564.1%
1Claude Sonnet 5.566.6%
3Claude Sonnet 558.9%

Ranks compare reported scores in the same published row; ties share a rank.

Protocols (understanding V2). §8.13.6: Benchling questions about applying/adapting real laboratory protocols. Haiku biology safeguards off.

Section 8, pages 111–139 of the Haiku 5.5 system card, read 8 October 2026. Values come from exact prose, tables and labelled figures. Unless specified, Haiku uses adaptive thinking at max effort and five trials with evaluation-specific context up to 1M. Each row retains its subset, metric, effort and source notes; no score is assigned to an unspecified SWE-Bench Pro version. Terminal runs use safeguards without fallback and no internet egress. ProgramBench uses filtered tasks and hidden-test pass rate without a time limit. OSWorld partial and strict results use the official 82-task offline subset. CAD uses 0–1 voxel IoU. Healthcare raw and length-adjusted scores stay separate. Life-science results disable Haiku biology safeguards. Independently run figures quoted by Anthropic remain publisher reports here. Missing comparison values are not inferred from rounded charts.

www.anthropic.com/claude-haiku-5-5-system-card

Sources

Every page used for this benchmark, read on 17 October 2026.