공고 • 6h
Unisound AI Technology Co., Ltd. Launches U2-Med Tri-Medical LLM
Unisound AI Technology Co., Ltd. completed the iterative upgrade of its medical large model based on the U2 foundation model and officially released the U2-Med Tri-Medical LLM, the industry’s first professional large model for the medical, medical insurance (including commercial insurance), and pharmaceutical sectors. The Upgrade further strengthens the Company’s model capabilities in professional fields and promotes the technological iteration of high-value Token scenarios. The model is expected to be widely embedded in the core processes related to the digital collaborative of tri-medical field, which will help accelerate the expansion of the Company’s large model business and generate long-term, sustainable revenue contributions. The U2-Med, relying on the Company’s U2 large model foundation, has achieved significant improvements in computational efficiency and intelligent agent execution. As the industry’s first industrial-grade large model to integrate the full-chain capabilities of the tri-medical, U2-Med incorporates the Company’s decade of deep experience in the medical field. It breaks down data and business silos across healthcare, medical insurance and pharmaceuticals, and supports the digital transformation of the tri-medical field with an integrated capability, comprehensively covering professional task scenarios across the broader medical sector. U2-Med delves deep into the core business processes of the tri-medical sectors, bringing tangible capability and efficiency improvements to clinicians, insurance practitioners, and pharmaceutical practitioners: for doctors, the model has capabilities such as assisted diagnosis and treatment, medical record generation, and identification of risks in clinical and electronic documents; for medical insurance and commercial insurance, the model can accurately identify questionable points such as use of medication beyond prescribed limits, duplicate charges, and excessive treatment, supports DRG/DIP compliance reviews, and helps audit personnel focus on key verifications; for the pharmaceutical industry, the model assists with pre-screening of trial subjects, verification of indications and contraindications, and risk alerts for concomitant medications, among other functions. This upgrade has truly achieved intelligent synergy across the tri-medical field, integrating clinical diagnosis and treatment, medical insurance payment, and pharmaceutical supply in a one-stop manner, while addressing the core pain points of the industry, including low cross-sector collaboration efficiency, fragmented control and high labour costs. In the July 2026 MedBench Evaluation, U2-Med models ranked first on the two leaderboards for full-modality large models and intelligent agents. It also leads in the Tri-Medical Comprehensive Benchmark for real-world business scenarios, achieving a comprehensive score of 89.1 in non-thinking mode and 88.9 in thinking mode. Its scores for specific tasks were 93.3 for medical record generation and quality control, 94.1 for post-diagnosis and medication management, and 88.8 for medical and commercial insurance tasks. The evaluation data fully demonstrates that U2-Med is not merely a general theoretical medical model, but a commercial-grade professional model that is aligned with real-world business processes. Its technical capabilities can be directly translated into tangible value for industrial deployment. Following the Upgrade, U2-Med will be deeply embedded into a wider range of tri-medical domain-wide business processes, creating higher industry value with its expert-level capabilities while replacing repetitive tasks to unleash the professional potential of tri-medical practitioners. The Company will also continue to promote its large-scale implementation in the tri-medical field, accelerating the positive flywheel of industry data accumulation and model iteration, continuously strengthening its business moat and injecting long-term momentum for the continuous release and commercial realization of high-value Tokens.