# HealthBench Professional Leaderboard

> Frontier language models ranked on HealthBench Professional, OpenAI's benchmark of 525 physician-authored tasks from real clinician conversations. Claude Fable 5 leads 9 models at 0.660. Updated 2026-08-16.

## Leaderboard

- [Leaderboard](https://healthbenchprofessional.com/): current ranking of all 9 models with scores, context windows, and API pricing
- [Full results, JSON](https://healthbenchprofessional.com/data/leaderboard.json): complete machine-readable results, CC BY 4.0
- [Full results, CSV](https://healthbenchprofessional.com/data/leaderboard.csv): one row per model
- [Plain-text leaderboard](https://healthbenchprofessional.com/llms-full.txt): the whole site's data in one document

## Background

- [What is HealthBench Professional](https://healthbenchprofessional.com/benchmark): how the benchmark was built and how it differs from HealthBench, Hard, and Consensus
- [Methodology](https://healthbenchprofessional.com/methodology): run protocol, grader configuration, comparability, limitations
- [FAQ](https://healthbenchprofessional.com/faq): direct answers on scores, grading, and the physician baseline

## Models

- [Claude Fable 5](https://healthbenchprofessional.com/models/claude-fable-5): 0.660, rank 1 of 9
- [GPT-5.6 Sol](https://healthbenchprofessional.com/models/gpt-5.6-sol): 0.605, rank 2 of 9
- [Claude Opus 5](https://healthbenchprofessional.com/models/claude-opus-5): 0.598, rank 3 of 9
- [Claude Sonnet 5](https://healthbenchprofessional.com/models/claude-sonnet-5): 0.578, rank 4 of 9
- [GPT-5.6 Terra](https://healthbenchprofessional.com/models/gpt-5.6-terra): 0.577, rank 5 of 9
- [Claude Opus 4.8](https://healthbenchprofessional.com/models/claude-opus-4.8): 0.558, rank 6 of 9
- [GPT-5.6 Luna](https://healthbenchprofessional.com/models/gpt-5.6-luna): 0.557, rank 7 of 9
- [GPT-5.5 Instant](https://healthbenchprofessional.com/models/gpt-5.5-instant): 0.384, rank 8 of 9
- [MAI-Thinking-1](https://healthbenchprofessional.com/models/mai-thinking-1): 0.350, rank 9 of 9

## Optional

- [Compare models](https://healthbenchprofessional.com/compare): head-to-head pages and the full score-difference matrix
- [Best AI for healthcare](https://healthbenchprofessional.com/best-ai-for-healthcare): picks by price tier
- [Updates](https://healthbenchprofessional.com/updates): dated changelog