Report
Claude Sonnet 5.5 and GPT-6 Astra lead new HealthBench and HealthBench Professional length-adjusted leaderboards, respectively
Artificial Analysis launched both leaderboards; a 2,000-character length adjustment changes some models’ rankings.
TLDR
Artificial Analysis’s Oct. 19 launch puts Claude Sonnet 5.5 first on HealthBench at 62.1% and GPT-6 Astra first on HealthBench Professional at 66.0% among published length-adjusted results. The benchmarks cover 5,000 health conversations and 525 clinician conversations, respectively. Replies longer than 2,000 characters lose points; on Professional, the adjustment moves raw-score leader Claude Opus 5.5 to sixth.
Combined views
—
12 Sources, first seen ago
