Which AI model scores highest on HealthBench?

GPT-5 by OpenAI holds the best HealthBench (hard) result among tracked models, at 46.2% (released Aug 7 2025). Higher scores are better on this benchmark.

What are the top 5 models on HealthBench?

1. GPT-5 (OpenAI) — 46.2%; 2. GPT-5 mini (OpenAI) — 40.3%; 3. GPT-6 Astra (OpenAI) — 34.2%; 4. gpt-oss-120b (OpenAI) — 30%; 5. GPT-6 Luna (OpenAI) — 25.4%.

How many models have a published HealthBench score?

7 tracked models have a published HealthBench (hard) score. Scores are collected from published lab reports and benchmark sources, with recorded sources credited alongside the results. We do not re-run the evaluations.

What is the best open model on HealthBench?

gpt-oss-120b by OpenAI is the highest-ranked model with downloadable weights on HealthBench (hard), scoring 30% at rank 4 overall.

Healthcare

HealthBenchhard

physician rubric score on the hard subset of the original HealthBench evaluation

full set full set, length-adjusted hard hard, length-adjusted consensus consensus, length-adjusted

Rankings

Higher is better

HealthBench — frequently asked questions

: physician rubric score on the hard subset of the original HealthBench evaluation
: GPT-5 by OpenAI holds the best HealthBench (hard) result among tracked models, at 46.2% (released Aug 7 2025). Higher scores are better on this benchmark.
: 1. GPT-5 (OpenAI) — 46.2%; 2. GPT-5 mini (OpenAI) — 40.3%; 3. GPT-6 Astra (OpenAI) — 34.2%; 4. gpt-oss-120b (OpenAI) — 30%; 5. GPT-6 Luna (OpenAI) — 25.4%.
: 7 tracked models have a published HealthBench (hard) score. Scores are collected from published lab reports and benchmark sources, with recorded sources credited alongside the results. We do not re-run the evaluations.
: gpt-oss-120b by OpenAI is the highest-ranked model with downloadable weights on HealthBench (hard), scoring 30% at rank 4 overall.