OpenAI says GPT-6 Astra was rated safe on 99% of 325 EHR tasks
Advancing computer-use agents for healthcare teams
AI summary
OpenAI publishes medical computer-use agent evaluation: GPT-6 Astra rated safe on 99% of 325 EHR tasks. On Stanford's HealthAdminBench, GPT-6 Astra scores 94.3%, GPT-6 Luna scores 91.8% at 8.5 cents per task. It also partners with Abridge, Cigna Healthcare, CVS Health and others.