YIDU TECH (02158) integrates DeepSeek-V4 into its MediGuide Intelligent Loop, enhancing the clinical support experience with dual model collaboration.

On April 24th, on the same day that DeepSeek officially released and open-sourced the new flagship model DeepSeek-V4 preview version, YIDU TECH (02158) announced that the clinical evidence-based intelligent body "Yidu intelligent circle" completed the medical scene adaptation of the dual-base architecture of DeepSeek-V4 and the Yidu large model. Million Token super-long context, stronger reasoning performance, and better ability to handle complex problems, three upgrades simultaneously landed in the three core scenes of "Yidu intelligent circle" clinical Q&A, medical record analysis, and evidence-based retrieval. In actual use, the system will flexibly call different model capabilities according to the difficulty of the question, with corresponding solutions for simple questions and complex tasks, and multiple models cooperating when necessary. In the past, a patient's complete medical record often contained dozens of pages of data such as admission records, daily progress notes, laboratory tests, imaging descriptions, and so on. Traditional AI was limited by the context window and could not read all at once, so doctors had to input in segments, leading to missed key clues. Now, DeepSeek-V4 has a 1 million token context window, equivalent to about 400,000 Chinese characters, enough to contain a complete hospitalization medical record or even an entire clinical guideline. Doctors only need to import the complete medical record once, for example, by asking directly, "What are the possible causes of septic shock in this patient," AI can refer to the entire progression from admission to present and provide an analysis based on global information. In the field of evidence-based medicine, traditional AI retrieval can only return relevant document fragments, and the comprehensive comparison across documents still needs to be manually completed by the doctor. Now, with the help of million contexts and stronger reasoning capabilities, "Yidu intelligent circle" can simultaneously integrate multiple clinical guidelines, original texts of randomized controlled trials, and case reports, automatically integrating evidence scattered across different sources, generating a complete evidence-based pathway, with each conclusion marked with its source, essentially pre-processing the literature review for the doctor. This upgrade brings three substantial changes: medical record analysis shifts from "breaking down input, fragment answers" to "complete input, global answers"; evidence-based retrieval shifts from "finding document fragments, doctors synthesizing themselves" to "comprehensive across documents, directly providing evidence-based pathways"; complex clinical problems shift from "shallow answers, long waits" to "more complete reasoning, more intelligent responses." Yidu intelligent circle is evolving from a tool to "help you check information" to a clinical workflow assistant that "helps you think." It is worth noting that along with technological upgrades and iterations, there often comes more prominent opportunities for large-scale implementation and market prospects, closely related to the strong business network of YIDU TECH in hospitals. According to public information, as of September 30, 2025, YIDU TECH had established in-depth cooperation relationships with 127 top domestic hospitals and 44 regulatory agencies, with the "AI medical brain" YiduCore as the core algorithm engine, covering more than 10,000 medical institutions in the hospital network. Among them, the solution based on the "Yidu intelligent circle" evidence-based engine has been implemented in more than 40 tertiary hospitals, deeply involved in over 500,000 clinical decision-making processes. With the accumulated cognitive abilities covering tens of thousands of diseases and knowledge graphs in core specialty areas such as tumors, blood diseases, and cardiovascular diseases, YIDU TECH has built a complete product matrix from "AI central platform" to "doctor Copilot" to "Yidu intelligent circle." Built upon benchmark large model technology, the cooperation cases of YIDU TECH with top domestic medical institutions further strengthen the company's high competitive barriers and broad commercial imagination space. For example, as the first national hospital to deeply integrate DeepSeek's "AI central platform," Xiangya Hospital of Central South University has successfully completed the localization deployment of the domestic AI medical central platform developed by YIDU TECH, creating an AI intelligent assistant that covers the entire diagnosis and treatment process. At the Cancer Prevention and Treatment Center of Sun Yat-sen University, YIDU TECH has achieved the private deployment of the AI central platform, linking the "YiduCore+DeepSeek" dual-engine with the hospital-wide big data platform to launch the "AI diagnosis and treatment assistant" covering all clinical scenarios, etc. Analysis indicates that with the deep coverage of over 10,000 hospital networks and the demonstrative driving effect of top benchmark hospital projects, YIDU TECH's integration of DeepSeek-V4 into the "Yidu intelligent circle" dual model architecture in this round not only achieves intergenerational improvement in core clinical evidence experiences, but also opens up great imagination space for subsequent large-scale commercialization the business flywheel constructed by top technology, top hospital collaboration network, and regional project capabilities is accelerating. The upgrade is now fully online, and doctors can experience it through the Yidu intelligent circle App or WeChat mini program.