All benchmarks
Research that holds up

Counts, lists and cross-references across permits, enforcement, grants, procurement and bill votes, such as facilities with an expired permit and an open violation.

Research across government datasets · overall

Exact answers

  • Justinian3.2
    96.9%
  • ChatGPTAstra 6.2
    43.8%
  • ClaudeFable 5.2
    34.4%
  • Grok4.7
    62.5%
  • Gemini4 Carbon
    26.7%
  • ClaudeOpus 5.5
    12.5%

What makes it hard

The answer sits in agency records, so it has to be computed from the data rather than reasoned out.

How it is scored

  • Exact answers: What share of answers match the figure or list computed from the source data?

This test measures access to government data as well as reasoning: each answer key is computed from the records.

Answers computed from official public datasets and rechecked against each publisher's own download or API.

Briefing

Test Justinian on your agency's own work.

Request a briefing →

This page’s interactive features did not load. Reloading usually fixes it. You can email sales@vulcan.ai for demos or support@vulcan.ai for help.