iFlytek Healthcare (02506) upgrades Spark Large Model V3.7, ranks first in MedBench5.0 evaluation, medical AI moves from "usable" to "easy to use"

On October 22, iFlytek Healthcare (02506) completed the upgrade of its Spark Healthcare Large Model V3.7, with multimodal capabilities covering the three major directions of speech, imaging, and semantics, achieving a leap from "the threshold of medical practicality" to "deep integration into the entire medical chain." In terms of specific applications, the diagnosis and treatment assistant for doctors has increased the end-to-end generation reasonableness rate in medical record writing scenarios to 85%, and its automatic report generation and quality control capabilities for complex medical imaging have reached industry-leading levels; iFlytek Xiaoyi for residents has added multimodal TCM consultation, dynamic memory management of health records, and personalized health plan generation, extending from one-time consultations to continuous health management. In terms of evaluation, the Spark Healthcare Large Model V3.7 ranked first in MedBench5.0 with a score of 99.6, outperforming mainstream large models at home and abroad on key tasks. What sustains the large model's continued leadership is the data flywheel accumulated from 1.71 million daily AI-assisted instances on the doctor side and 520,000 patient-side AI services.