A rubric-graded benchmark of frontier language models on current clinical evidence.

Leaderboard

v1 release set, 89 tasks. Updated August 16, 2026.

83.8

81.3

78.3

77.9

77.8

68.4

57.9

53.9

50.1

35.5

33.0

29.2

14.4

7.8

TM

Claude Fable 5

Grok 4.6

Claude Opus 5

GPT-5.6 Sol (max)

Kimi K3

Muse Spark

Gemini 3.6

Inkling

Claude Sonnet 5

MiniMax M3

MAI Thinking

GLM 5.2

Mistral Medium 3.5

Nemotron 3.5 Lightning

Consensus rubric credit on the release set, one answer per task, tools disabled. 95 percent bootstrap confidence intervals over tasks accompany every score on the full ranking. GPT-5.6 Sol (high) is completing its run and will be added when grading finishes.

Health Optimization Bench evaluates language models on hard, freshness-dependent questions in preventive and optimization medicine. Every task is written against a primary source, audited by model families that did not author it, and scored blind by a panel of three independent families. The task's authoring family never grades it.

Micro benches

The bench is assembled from focused micro benches, each covering one area of preventive and optimization medicine. The incretin therapeutics evidence suite is the first with a released model ranking. The remaining suites are authored and are moving through the same cross-family verification pipeline; clinical decision tasks are additionally reviewed by licensed clinicians before release.

• Incretin Therapeutics: Evidence Synthesis
  89 tasks · ranking live
• Incretin Therapeutics: Clinical Decisions
  109 tasks · under clinician review
• Cancer Screening & Early Detection
  13 tasks · in verification
• Exercise & Cardiorespiratory Fitness
  15 tasks · in verification
• Longevity / Geroscience Pharmacology
  15 tasks · in verification
• Hormone Optimization
  15 tasks · in verification
• Blood Pressure Optimization
  15 tasks · in verification
• Lipid & ASCVD Prevention
  15 tasks · in verification
• Nutrition & Supplements
  15 tasks · in verification
• Sleep Optimization
  14 tasks · in verification

Data sample

A 30 task sample, drawn from each micro bench, is available through the Arcophos research data platform, with rubric criteria and reference responses included.

Paper

The methodology paper is in preparation and will accompany the archived v1 release, together with per-task verification records and the grading protocol. For early access, licensing, or evaluation of a private model, contact Arcophos.