UNISOUND (09678) announced the release of its latest medical expert large model, "Shan Hai Zhi Yi 5.0" ("Medical Large Model"). This version establishes a dual-core product system comprising a "Medical Text Large Model" and a "Medical Multimodal Large Model," comprehensively covering core capabilities such as text processing, intelligent agent collaboration, and multimodal perception. The model integrates diverse functions, including medical knowledge Q&A, clinical task planning, and medical imaging analysis, forming one of the industry's most comprehensive AI-powered medical technology support systems.

The Medical Large Model is capable of deep clinical reasoning simulation, enabling precise differential diagnosis and evidence-based decision-making. Leveraging extensive high-quality medical data and proprietary knowledge bases, UNISOUND has achieved top-tier performance in core tasks such as medical Q&A and information extraction. Through continuous curriculum learning, knowledge graph enhancement, and evidence-based alignment training, the model minimizes medical hallucinations while ensuring output accuracy and verifiability.

In December 2025, under the MedBench 4.0 evaluation framework—a leading Chinese medical large model benchmark developed by the Shanghai AI Research Institute—UNISOUND scored 94.6, 65.7, and 59.6 in medical intelligent agents, large language models, and multimodal large models, respectively, securing first place across all three technical paradigms. Additionally, the model was recognized in the "MedAIBench Ranking (Top Domestic Medical Large Models)" report released by the National AI Application Pilot Base (Healthcare) in Zhejiang, showcasing its robust AI-driven medical expertise.

UNISOUND's Medical Large Model has already been deployed at scale in numerous top-tier hospitals. It has undergone rigorous technical validation in high-demand clinical scenarios, such as assisted diagnosis and treatment, demonstrating deep alignment with medical industry standards and real-world clinical needs—further solidifying its practical value and industry applicability.