UNISOUND (09678.HK): Self-Developed Medical Imaging Multimodal Large Model U2-RadiMed Officially Released, Ranking First in Comprehensive Benchmarks

NewTimeSpace News: On 13 August 2026, Unisound AI Technology Co., Ltd. (stock code: 09678) voluntarily announced that, following the release of the U2-Med three-medicine large model, the Company has recently officially released its self-developed medical imaging multimodal large model "U2-RadiMed". The model is the first full-chain imaging large model in China that deeply integrates medical imaging understanding, report generation, clinical reasoning and decision-making, and visual question answering, supporting multimodal mixed input of single or multiple images and text, and comprehensively covering mainstream medical imaging modalities including CXR (chest X-ray), CT and MRI.

According to the announcement, U2-RadiMed establishes a framework of six core capabilities: report generation, lesion detection and localisation, clinical reasoning, clinical decision-making, general medical VQA and advanced medical reasoning, enabling a complete closed loop of "reading images, retrieving knowledge, reasoning and drawing conclusions". In multiple authoritative horizontal benchmarks, the model ranked first in the general comprehensive assessment with a composite score of 55.47, ahead of domestic and international models including Gemini 3.1 and GPT-5.4; in the special assessments of the six core capabilities, it swept first place in four categories - report generation, lesion detection and localisation, general medical VQA and advanced medical reasoning - with lesion detection and localisation scoring 37.58, far ahead of comparable models; and it ranked first among participating models in the MedBench 5.0 all-modality large model assessment with a composite score of 57.2.

At the application level, U2-RadiMed breaks through the limitation of traditional AI single-image analysis, capable of understanding the correlation logic between multi-temporal images and supporting diverse clinical scenarios including lesion detection and localisation, multi-temporal image comparison, dynamic disease course tracking, differential diagnosis and assisted report generation, providing clinicians with comprehensive AI-assisted support from lesion screening and precise diagnosis to individualised treatment decision-making. The model has been launched on the Company's MaaS platform (Token Hub), supporting multi-image input and a 40K context window, with out-of-the-box model capabilities that can be rapidly deployed to imaging departments and clinical diagnosis and treatment scenarios of medical institutions at all levels.

The Company stated that the release of U2-RadiMed marks its large model capabilities formally entering the more complex and professional core clinical scenario of medical imaging, further improving its medical AI product system of "general foundation large model - three-medicine professional large model - medical imaging multimodal large model". Going forward, the Company will continue to deepen the application and innovation of large models in the broader medical field, driving medical imaging AI from technological breakthroughs towards real medical productivity.

NewTimeSpace Disclaimer: All content herein is the original work of NewTimeSpace. Any reproduction, reprinting, or use of this content in any other manner must clearly indicate the source as "NewTimeSpace". NewTimeSpace and its authorized third-party information providers strive to ensure the accuracy and reliability of the data, but do not guarantee the absolute correctness thereof. This content is for reference only and does not constitute any investment advice. All transaction risks shall be borne by the user.