Health AI Benchmarks

Anthropic

11 models with published results, newest first. Anthropic website.

Health Evals Index: this site's own calculation from 4 boards; “N of 4” shows how many each score rests on. How it works

Scroll for benchmarks

Models with their index score and rank on each board
ModelReleasedIndex, with boards behind itHealthBench ProfessionalHealthBenchHealth Optimization Bench: incretin evidence synthesisArtificial Analysis Healthcare & Medical IndexHealth Optimization BenchSAFE-Teen
Claude Haiku 5.57 Oct 202660.3 #102 of 4 boards6/437/33
Claude Sonnet 5.528 Sep 202664.0 #23 of 4 boards2/433/332/18
Claude Opus 5.522 Sep 202664.0 #13 of 4 boards4/431/331/296/18
Claude Fable 5.11 Sep 202661.7 #63 of 4 boards12/4310/331/182/298/16
Claude Opus 524 Jul 202663.1 #43 of 4 boards17/432/335/183/292/16
Claude Sonnet 530 Jun 202656.8 #253 of 4 boards19/4312/3313/1811/16
Claude Fable 59 Jun 202663.4 #33 of 4 boards3/436/332/184/291/16
Claude Opus 4.828 May 202657.9 #172 of 4 boards21/4311/33
Claude Opus 4.716 Apr 202654.9 #351 of 4 boards30/43
Claude Sonnet 4.617 Feb 202651.9 #421 of 4 boards38/43
Claude Haiku 4.515 Oct 202547.1 #481 of 4 boards43/43