The grounding number: HealthBench Professional 63.4%
GPT-6: Astra scores 63.4% on HealthBench Professional, OpenAI's evaluation of professional-grade medical reasoning. That leaves more than a third of professional-grade healthcare tasks unresolved even on OpenAI's own harness — treat it as a capable assistant for a licensed professional, not a diagnostic tool.
Example prompt
Summarize the attached de-identified clinical note into a structured problem list, medication list, and open questions for physician review. Do not suggest a diagnosis or treatment change; flag anything that requires clinician judgment.Who should look elsewhere
If you need a certified medical device, diagnostic decision support cleared for clinical use, or handling of protected health information without a completed compliance and data-processing review, GPT-6: Astra is not that product on its own and you should look for a purpose-built, regulator-cleared system instead.
Ready to test the workflow?
Recommended access
Use pay-as-you-go API credits for a controlled evaluation on your own workload and cost profile before committing production routing.