Can AI read a chest X-ray as well as a radiologist?

RadLE was built to answer a blunt question that most healthcare AI vendors avoid: how far are current frontier models from expert clinical performance when the cases are messy, ambiguous, and real?

The benchmark compares expert radiologists, trainees, and multimodal AI models on chest radiology tasks that require clinical reasoning rather than pattern matching alone.

Its headline result matters beyond rankings: the best AI crossed trainee-level performance, but still remained far behind experts, defining a clearer threshold for responsible deployment.