New HealthBench leaderboards compare AI models’ replies to health conversations

ArtificialAnlys launched leaderboards for two OpenAI benchmarks: HealthBench covers 5,000 conversations with everyday users and health workers, while HealthBench Professional covers 525 clinician conversations. It reports GPT-6 Astra (max) at 66.0% and GPT-6.1 Sol (max) at 65.1% on Professional’s length-adjusted score. Claude Sonnet 5.5 (max with fallback) led HealthBench at 62.1%. The adjustment lowers scores for replies over 2,000 characters and raises them for shorter replies.