OpenAI says its new ChatGPT for Clinicians outperforms doctors on clinical tasks even when they have unlimited time and web access

The benchmark design matters more than the headline number: 'unlimited time and web access' sounds like a generous condition for physicians, but it also describes an artificial test environment that strips out the cognitive load, interruptions, and incomplete information that define actual clinical work. OpenAI is grading on a curve it drew itself.

OpenAI's recent moves suggest a company expanding aggressively across verticals rather than consolidating any single one. The April 17 TechCrunch piece on OpenAI's acquisition spree noted the company was buying into finance and media simultaneously, and ChatGPT for Clinicians fits that same pattern: plant a flag, generate coverage, iterate later. Meanwhile, the MIT Technology Review piece from April 16 made the sharper point that benchmark performance is largely irrelevant to enterprise deployment — what matters is the operational infrastructure around the model, including liability, audit trails, and integration with clinical workflows. None of that appears in OpenAI's announcement.

Watch whether any independent clinical institution publishes a prospective evaluation of ChatGPT for Clinicians against the same benchmark within the next six months. If no third-party replication appears, the performance claims should be treated as marketing until proven otherwise.