Data

Name: HealthBench Hard leaderboard results
Creator: Arcophos
License: https://creativecommons.org/licenses/by/4.0/

Everything the leaderboard shows, every score and every API fact, exists as machine-readable files at URLs that never move. Files and pages regenerate together and share the 2026-08-16 version stamp, and the CC BY 4.0 license means reuse just requires attribution.

Downloads

leaderboard.json
the complete results, benchmark metadata included
leaderboard.csv
spreadsheet-ready, one model per row
llms.txt
the site indexed for language model agents
llms-full.txt
every number on the site in one plain-text file

Citing this leaderboard

A suggested format:

Arcophos. HealthBench Hard Leaderboard. August 16, 2026 snapshot. https://healthbenchhard.ai. Accessed [date].

The benchmark itself is cited via OpenAI's paper at arxiv.org/abs/2505.08775. Numbers here come from the runs described on the methodology page and will not always match the paper's.

Versions

2026-08-16

First release, 9 models. This is the current file.

Download URLs are permanent. Fields get added, never renamed. Every change is announced on the updates page.