Skip to content

OfficeQA

Grounded numerical reasoning over a corpus of historical U.S. Treasury Bulletin documents.

3 models

Higher is better.

  1. 1Claude Opus 5.5Anthropic78.9
  2. 2Claude Opus 5Anthropic78.1
  3. 3Claude Sonnet 5.5Anthropic76.9
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Agentic grounded QA accuracy
Tasks
Historical Treasury Bulletin questions
Authors
Databricks and Anthropic (2026)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.