OpenAI Releases MentalHealthBench to Evaluate Frontier Models in Mental Health Conversations

OpenAI has introduced MentalHealthBench, an open benchmark for testing how frontier AI models handle realistic mental health conversations, saying it was built with input from more than 80 mental health clinicians. The release gives outside researchers a new way to examine model behavior in sensitive, multi-turn situations where safety, empathy and clinical judgment can be difficult to score.