Hippocratic AI has announced the launch of Polaris 3.0, a suite of 22 large language models (LLMs) with 4.2 trillion parameters, designed to enhance patient safety and engagement in non-diagnostic clinical tasks. This latest version builds on real-world patient and provider feedback, improving clinical accuracy to 99.38% and increasing patient satisfaction scores. Key advancements include deep-thinking models for error reduction, improved clinical documentation accuracy, enhanced emotional intelligence, and robust audio handling features. Additionally, Polaris 3.0 strengthens multilingual safety capabilities, workflow integration with electronic medical records (EMRs), and orchestration features to better navigate healthcare operations.
Munjal Shah, Co-founder and CEO of Hippocratic AI, emphasized the rigorous real-world testing behind Polaris 3.0, highlighting its unmatched safety and effectiveness in patient-facing clinical operations. The company has incorporated insights from over 1.85 million patient calls and testing by more than 6,200 licensed clinicians. Alongside this release, Hippocratic AI introduced a novel Real World Evaluation of Large Language Models in Healthcare (RWE-LLM) framework to guide AI safety in the industry. Polaris 3.0 also brings enhancements in patient engagement metrics, reducing AI refusal rates and improving call completion rates, solidifying its position as a leading vertical AI solution in healthcare.