iFLYTEK Healthcare (02506) upgraded the Spark Big Model to V3.7, leading the MedBench 5.0 evaluation. Medical AI moved from “usable” to “easy to use”

The Zhitong Finance App learned that on October 22, iFLY Healthcare (02506) completed the V3.7 upgrade of Spark Healthcare's big model. Its multi-modal capabilities covered the three major directions of voice, video, and semantics, making the leap from “medical practical threshold” to “deep integration into the entire medical chain”.

Specifically, diagnosis and treatment assistants for physicians have increased the reasonable rate of end-to-end generation of medical record writing scenarios to 85%, and the automatic report generation and quality control capabilities of complex medical images have reached the industry's leading level; iFLYX has added multi-modal traditional Chinese medicine consultation, dynamic memory management for health files, and personalized health plan generation for residents, extending from a single consultation to continuous health management.

In terms of evaluation, Spark Healthcare's big model v3.7 ranked first with a score of 99.6 in MedBench 5.0. The key tasks are superior to mainstream models at home and abroad. What supports the model's continued lead is a data flywheel with 1.71 million doctor-side AI assistants and 520,000 patient-side AI services every day.