Yun Zhisheng (09678) released self-developed multi-modal large medical imaging model “U2-radimed”

Zhitong Finance App News, Yun Zhisheng (09678) announced. Following the release of the U2-Med Three Medical University models, the company recently officially released the self-developed multi-modal medical imaging model “U2-radimed”.

The model is the first full-link imaging model in China that deeply integrates medical image understanding, report generation, clinical reasoning decision-making, and visual question-and-answer. It supports multi-modal mixed input of single and multiple images and text, fully covering mainstream medical imaging types such as CXR (chest X-ray), CT, and MRI, and constructs a framework of six core competencies: report generation, lesion detection and localization, clinical reasoning, general medicine VQA, and advanced medical reasoning. It can complete a complete closed loop of “viewing images → adjusting knowledge → making inferences → drawing conclusions”. In many authoritative horizontal evaluations, U2-radimed ranked first in the general comprehensive evaluation with a comprehensive score of 55.47, leading domestic and foreign models such as Gemini 3.1 and GPT-5.4; in the six core competency assessments, it ranked first in the four categories of report generation, lesion detection and localization, general medicine VQA, and advanced medical reasoning. Among them, lesion detection and localization led similar models by a score of 37.58; and ranked first among the participating models with a comprehensive score of 57.2 in the MedBench 5.0 full modal large model evaluation.

At the application level, U2-radimed breaks through the limitations of traditional AI single image analysis, can understand the correlation logic between multi-time series images, supports multiple clinical scenarios such as lesion detection and positioning, multi-time image comparison, dynamic disease tracking, differential diagnosis, and report assisted generation, and provides clinicians with a full range of AI support from lesion screening and accurate diagnosis to individualized treatment decisions. Currently, the model has been launched on the company's MaaS platform (Token Hub), which supports multiple image input and 40K contexts, enables the model's ability to be used right out of the box, and can be quickly connected to imaging departments and clinical diagnosis and treatment scenarios of medical institutions at all levels.

The release of U2-radimed marks the official entry of the company's large model capability into the more complex and professional clinical core scenario of medical imaging, further improving the company's medical AI product system of “General Pedestal Big Model — Three Medical Professional Large Model — Medical Imaging Multi-modal Large Model”, and consolidating the company's leading position in the field of medical artificial intelligence. In the future, the company will continue to deepen the application and innovation of big models in the pan-medical field, promote medical imaging AI from technological breakthroughs to true medical productivity, help medical institutions improve quality and efficiency, and create long-term value for the company's shareholders.