AIEvals

Benchmarks

OfficeQA

Answering grounded questions over a large corpus of historical United States Treasury Bulletin documents in an agentic workflow.

Document reasoningReported by AnthropicVendor-reported onlyReported by Anthropic

How it is run

Section 8.10.1 preselects the relevant documents and supplies extracted text, reporting a mean across five runs at max effort.

What it cannot tell you

Document retrieval is preselected in this configuration. OfficeQA Pro is a harder 133-question subset and is reported separately.

Why there is no independent table

No independent suite in this dataset runs OfficeQA. 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.

  • OfficeQA ProSame family. Grounded reasoning on the harder 133-question OfficeQA Pro subset of historical Treasury Bulletin documents.
  • GDP.pdfSame category. Reasoning over long professional PDF documents, requiring all task criteria to pass.

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 5.573.5%
3Claude Haiku 4.563.0%
1Claude Sonnet 5.576.9%

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

OfficeQA. §8.10.1: preselected documents as extracted text, mean of five max-effort runs. Haiku 4.5 fixed 128k thinking budget.

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.