All Benchmarks/Knowledge

HealthBench Professional

Often citedHigher is better

OpenAI's test for medical professionals: responses to 525 tasks close to clinical work are graded against rubrics. Values here are standardized on the length-adjusted score. Scores run from 0 to 100%. Higher is better.

Other versions:HealthBenchHealthBench Hard

This ranking is built only from scores the model makers published themselves. Test settings such as tool use and reasoning effort differ by maker, so check each score's setting before comparing models directly.

Top score65.6%Claude Opus 5.5
Models tested21
Last updated2026.09.22
Model release date (newest on the right)
Best score so farHigher on the chart is better
  1. 1
    Anthropic
    Claude Opus 5.5Self-reported
    2026.09.22 · length-adjusted · no tools · max (adaptive thinking)
    65.6%
  2. 2
    OpenAI
    GPT-6 AstraSelf-reported
    2026.09.04 · length-adjusted
    64.7%
  3. 3
    Anthropic
    Claude Fable 5Self-reported
    2026.06.09 · length-adjusted · no tools
    63.3%
  4. 4
    Anthropic
    Claude Fable 5.1Self-reported
    2026.09.01 · length-adjusted · no tools · max (adaptive thinking)
    62.1%
  5. 5
    OpenAI
    GPT-6 LunaSelf-reported
    2026.09.22 · length-adjusted
    60.8%
  6. 5
    OpenAI
    GPT-6 SolSelf-reported
    2026.09.22 · length-adjusted
    60.8%
  7. 7
    OpenAI
    GPT-5.6 SolSelf-reported
    2026.07.09 · length-adjusted
    60.5%
  8. 8
    Anthropic
    Claude Opus 5Self-reported
    2026.07.24 · length-adjusted · no tools · max (adaptive thinking)
    59.8%
  9. 9
    Meta
    Muse Spark 1.1Self-reported
    2026.07.09 · length-adjusted · xhigh
    59.3%
  10. 10
    Anthropic
    Claude Sonnet 5Self-reported
    2026.06.30 · length-adjusted · no tools · max (adaptive thinking)
    57.8%

The "harness" is the agent program the AI used to carry out the task. The same model can score very differently depending on its harness and reasoning effort.

Vendor-reported numbers use varying setups — treat them as indicative.