Yes, AI models have scored at or above passing thresholds on USMLE Step 1 practice exams. One widely cited evaluation put a large language model at 86%, with no medical school, clinical rotations, or supervised patient care.