OpenAI Launches Free ChatGPT for US Clinicians — Claims It Outperforms Doctors on New HealthBench
OpenAI launched a free, verified-access ChatGPT tier for US physicians, NPs, PAs, and pharmacists, powered by GPT-5.4. The tool includes clinical search across peer-reviewed literature and repeatable workflow skills for referral letters and patient instructions. OpenAI simultaneously open-sourced HealthBench Professional — a new clinical benchmark — and claims ChatGPT outperforms specialty-matched physicians with unlimited time on it.
Original sourceOpenAI launched ChatGPT for Clinicians on April 23, 2026 — a free, verified-access version of ChatGPT built specifically for US physicians, nurse practitioners, physician assistants, and pharmacists. Verification is handled through existing healthcare credentialing systems, and access is being rolled out in waves to prevent abuse.
The product is powered by GPT-5.4 and adds several clinical-specific features not in the standard ChatGPT interface. Clinical Search pulls from peer-reviewed literature, clinical guidelines, and drug databases rather than the open web. Repeatable workflow "skills" let clinicians build template flows for common documentation tasks — referral letters, patient discharge instructions, differential diagnosis frameworks — and reuse them across cases.
OpenAI's headline claim is that ChatGPT scores higher than specialty-matched physicians on the new HealthBench Professional benchmark, even when those physicians have unlimited time and web access. That's a significant statement, and the company is inviting the research community to verify it by simultaneously open-sourcing the HealthBench Professional benchmark under a permissive license.
The clinical AI space is increasingly crowded — Microsoft/Nuance's DAX Copilot, Google's MedPaLM 2 descendents, and a wave of VC-backed startups are all targeting the same workflow. OpenAI's play is to use the free tier as acquisition, convert clinicians to paid plans for advanced use, and establish GPT as the default clinical reasoning layer before competitors entrench.
For healthcare privacy advocates, questions remain: the product requires patients' information to pass through OpenAI's infrastructure, and BAA agreements for HIPAA compliance are only available on the Team and Enterprise plans — not the free clinician tier. OpenAI says PHI should not be entered in the free version, which raises practical questions about how useful the tool is for real clinical documentation without patient data.
Panel Takes
The Builder
Developer Perspective
“Open-sourcing HealthBench Professional alongside the launch is the right move — it lets the community stress-test the 'outperforms doctors' claim independently. If the benchmark holds up, this is a major milestone. If it's narrow or gamed, we'll know within weeks.”
The Skeptic
Reality Check
“The HIPAA gap is a dealbreaker for actual clinical documentation — you can't have a useful clinical AI that can't touch patient data. Free tier with no BAA is a trial product, not a clinical tool. Real adoption will depend on whether hospitals will pay Enterprise pricing for the compliant version.”
The Futurist
Big Picture
“Healthcare is the highest-stakes domain AI is entering and this is OpenAI's most explicit claim that AI reasoning can match specialist physicians. If true, the implications for rural care access, healthcare costs, and the physician training pipeline are profound. This is the most important AI launch of the week.”