Hippocratic AI announced Polaris 5.0, a new version of its healthcare-focused AI model designed for real-world patient voice interactions. Built on more than 180 million patient interactions and validated by thousands of U.S.-licensed clinicians, the model was tested against leading systems from companies like OpenAI, Anthropic, and Google. Across clinical accuracy, safety, compliance, empathy, and conversational performance, Polaris 5.0 outperformed these models, while also operating fast enough for live voice conversations. The system uses a large “constellation” architecture with a 700-billion-parameter core and added capabilities like speech recognition tuned for medical language, drug safety checks, and real-time conversational handling.
The update introduces more than 20 new healthcare skills, including clinical escalation, insurance and benefits understanding, multilingual support, and improved emotional awareness during patient conversations. It can detect symptoms like coughing during calls, navigate external systems such as pharmacy menus, and handle complex tasks like cross-referencing medical and insurance documents in a single interaction. Polaris 5.0 also shows high performance in safety and compliance benchmarks, including near-perfect accuracy in drug safety and clinical escalation, while maintaining fast response times. Overall, the release positions the model as a purpose-built, safety-focused AI system designed to support patient-facing healthcare workflows at scale.