LLM Reference
activeMedicalSafety

HealthBench Hard

Metric: Length-adjusted score (higher is better)Introduced: 2025

Hard subset of OpenAI HealthBench for more difficult medical conversations. LLMReference uses the length-adjusted score for cross-model comparison.

Models ranked

3

tracked on this benchmark

Score band

33.1 – 32.0

best → lowest tracked

Snapshot trend

—

need ≥2 snapshots

Leaderboard

Tracked models ranked by Length-adjusted score (higher is better).

Compare candidates
#Model variant and provenanceScore
1
GPT-5.6 Sol
Version: length-adjusted; raw 31.1; mean response 1751 charsHarness: Not recordedEvaluator: Not recordedObserved: Jun 26, 2026Confidence: Not recordedSource

Notes: Official GPT-5.6 preview system-card row. Raw score is recorded only in the version label so rankings use the comparable length-adjusted metric.

33.1
2
GPT-5.6 Terra
Version: length-adjusted; raw 34.3; mean response 2199 charsHarness: Not recordedEvaluator: Not recordedObserved: Jun 26, 2026Confidence: Not recordedSource

Notes: Official GPT-5.6 preview system-card row. Raw score is recorded only in the version label so rankings use the comparable length-adjusted metric.

32.7
3
GPT-5.6 Luna
Version: length-adjusted; raw 31.4; mean response 1923 charsHarness: Not recordedEvaluator: Not recordedObserved: Jun 26, 2026Confidence: Not recordedSource

Notes: Official GPT-5.6 preview system-card row. Raw score is recorded only in the version label so rankings use the comparable length-adjusted metric.

32.0

How to read this benchmark

This benchmark scores models where higher is better. Use scores for directional filtering and shortlisting, not universal quality ranking; then validate pricing, context window, provider availability, and fit for your workload.

Trust this score when

  • There is a fresh timestamped snapshot (or multiple snapshots) for this benchmark.
  • The model list covers the same version family you can actually deploy today.
  • Top candidates overlap with your required routing and feature requirements.

Be cautious when

  • There is only one benchmark snapshot or the dataset appears stale.
  • The benchmark metric direction is opposite of your decision objective.
  • The score difference between options is narrow and likely within implementation variance.

Related benchmarks

Last reviewed: Jun 26, 2026