2025 年医疗大模型品牌推荐 2025 年医疗大模型品牌推荐 海量知识深度整合，智能生成革新医疗范式 | 一、市场背景 | 2 | | --- | --- | | 1.1 摘要 | 2 | | 1.2 医疗大模型定义 | 2 | | 1.3 市场演变 | 2 | | 二、市场现状 | 3 | | 2.1 市场规模 | 3 | | 2.2 市场供需 | 3 | | 三、市场竞争 | 3 | | 3.1 市场评估维度 | 3 | | 3.2 市场竞争格局 | 4 | | 3.3 十大品牌推荐 | 4 | | 四、发展趋势 | 5 | | 4.1 底层算法与数据质量升级 | 6 | | 4.2 应用形态与部署模式多元化 | 6 | | 4.3 商业模式与生态构建平台化 | 6 | 自 2019 年医疗大模型诞生以来，其发展经历了从通用模型的基础能力迁移，到专业医 疗知识的深度对齐，再到多模态与临床工作流融合的快速演进。早期模型如 BioBERT、ClinicalBERT 通过生物医学语料预训练夯实了基础能力；随后 Med-PaLM、 HuaTuoGPT 等模型通过指令微调与知识增强，在医学问答、报告生成等任务中展 ...

| 。大语言模型评测榜单 | | 壹看更多 | ~ 多模态大模型评测榜单 | 查看更多 > | 2 智能体评测榜单 | 查看更多 | | --- | --- | --- | --- | --- | --- | --- | | 20 | 华为云健康管理大模型 | 71.0 | 数坤坤多模态医学大模型V3 83 | 63.6 | 惠每医疗大模型 Pas | 95.5 | | | Huawei | | 数坤科技 | | 上海新创惠每科技有限公司 | | | 127 | 千问健康大模型 | 70.8 | 微医医疗大模型 25 | 60.8 | UniGPT-Med-U1 (2) | 94.6 | | | 千间C端 | | 微医 | | 云知声智能科技股份有限公司 | | | 37 | WiseDiag v2 | 69.8 | UniGPT-Med-VL 735 | 59.6 | WiseDiag v2 R35 | 94.3 | | | 杭州智诊科技有限公司 | | 云知声智能科技股份有限公司 | | 杭州智诊科技有限公司 | | | | 微医医疗大模型 | 68.2 | GPT-5-chat-latest A | 5 ...

Core Viewpoint - JD Internet Hospital is launching the "New Year Health Guard" initiative to provide 24-hour online consultation services, integrating resources from top-tier hospitals to address health needs during the Spring Festival [1] Group 1: Service Overview - The initiative aims to alleviate healthcare access issues during the holiday season, ensuring timely and professional medical support for users [1] - Users can access the service through the JD APP by searching for "New Year Health Guard" or "Ask a Doctor," allowing for quick entry into the consultation channel [1] Group 2: User Experience - Previous experiences highlight the effectiveness of JD Internet Hospital's online consultation, such as a case where a user abroad received timely advice for a burn injury through video consultation with a top-tier hospital [3] - Another case involved a 71-year-old user who received a diagnosis and treatment plan for severe skin itching, demonstrating the platform's ability to provide effective medical guidance [3] Group 3: AI Integration - JD Health's AI doctor, "Dai Wei," operates around the clock, utilizing a self-developed medical model and comprehensive medical knowledge base to offer multi-round symptom collection, analysis, and recommendations [4] - The AI doctor integrates authoritative clinical guidelines and medical literature, providing evidence-based suggestions to enhance the quality of health services [4] Group 4: Strategic Goals - JD Health aims to leverage its internet medical service advantages and robust pharmaceutical supply chain capabilities to ensure quality medical resources are accessible regardless of users' locations [4] - The initiative seeks to provide peace of mind and health support for families during the reunion moments of the Spring Festival [4]

Core Insights - The medical large model sector is experiencing significant advancements, particularly with the recent launches from OpenAI, Anthropic, and Google, indicating a competitive landscape in healthcare AI [1][3][4] - Domestic companies in China, such as Baichuan Intelligence and Ant Group, are making substantial progress, showcasing their capabilities and innovations in the medical AI field [5][6][10] Group 1: OpenAI's Developments - OpenAI launched ChatGPT Health and OpenAI for Healthcare, which comply with HIPAA regulations and aim to enhance patient care services [2] - ChatGPT Health integrates a new module based on the latest GPT-5 model, allowing users to connect personal health data while ensuring data privacy [2] - OpenAI for Healthcare targets medical institutions, offering enterprise-level solutions for clinical environments, including automated documentation [2] Group 2: Competitors' Responses - Anthropic released Claude for Healthcare, directly competing with OpenAI's offerings, highlighting its commitment to the healthcare sector [3] - Google updated its open-source medical large model, MedGemma 1.5, improving its capabilities in text, medical records, and medical imaging [3] Group 3: Domestic Innovations - Baichuan Intelligence launched Baichuan-M3, surpassing GPT-5.2 in benchmark tests, with a hallucination rate of only 3.5%, which was later improved to 2.6% [4][5] - Ant Group's medical model has gained significant traction, achieving 30 million monthly active users and doubling daily inquiries within a month [5] - JD Health and other domestic firms are also releasing new models and solutions, contributing to a vibrant competitive landscape in China's medical AI sector [5][10] Group 4: Benchmarking and Performance - HealthBench and MedBench are the primary benchmarking systems for evaluating medical large models, assessing safety, clinical applicability, and professional reasoning [7][8] - The performance of medical large models is crucial, with ongoing iterations and resource investments being necessary for advancements [8] Group 5: Industry Trends and Future Outlook - The medical large model sector is expected to grow significantly, with various domestic models emerging as strong contenders in 2026 [6][10] - The integration of local clinical guidelines and data security measures positions domestic models as more suitable for the Chinese market [6] - The increasing support from the government and the focus on localized solutions are likely to enhance the competitiveness of domestic models [6]

Group 1 - The core viewpoint is that AI in healthcare is integrated into national strategic planning, leading to an accelerated phase of industry development through a comprehensive policy framework [1][6] Group 2 - Infrastructure support focuses on standardization to enhance data interoperability and smart hospital construction, laying the groundwork for AI applications in healthcare [2] - Regulatory governance has established a tiered framework to balance innovation and safety, providing clear guidelines for AI medical software classification and lifecycle regulation [3] - Payment mechanism reforms are crucial for the commercialization of AI healthcare products, with policies supporting AI-assisted diagnosis and innovative payment models [4] Group 3 - Application scenarios are being expanded, with the government identifying key areas for AI in healthcare, creating a positive cycle of policy guidance and innovation [5] - The integration of AI with pharmaceutical manufacturing and healthcare services is encouraged, fostering a diversified innovation ecosystem [6] Group 4 - Shanghai is leading the way in medical AI development with the first provincial-level plan, aiming to create a nationally replicable model for AI in healthcare [7] - The plan emphasizes enhancing innovation sources, building support platforms, and creating comprehensive application scenarios in clinical and public health [8][9][10] Group 5 - Companies are focusing on three critical elements: application scenarios, data, and computing power, which are essential for breakthroughs in the AI healthcare industry [11][12] - There is a pressing need for high-value application scenarios, as many AI technologies face challenges in adapting to clinical and industry needs [13][14] - Data sharing and quality are significant concerns, with companies advocating for unified data-sharing mechanisms to overcome barriers [16][17] - The demand for computing power is high, and companies are calling for centralized computing resources to reduce costs and improve efficiency [17] Group 6 - The trend of companies expanding internationally is becoming essential, with different paths for AI in drug development and medical devices [18] - AI in drug development is characterized by a dual approach of licensing out and independent international expansion, focusing on regions with established clinical trial systems [19][20] - AI in medical devices is pursuing multiple strategies to adapt to global market needs, including remote surgery and tailored solutions [20] Group 7 - A comparison of capabilities between Chinese and American companies in various AI healthcare fields shows differences in market maturity and regulatory environments [30] Group 8 - Specialized medical models are gaining traction, with hospitals leading the deployment of AI models tailored to specific diseases [34] - Evidence-based medicine is being utilized to address challenges in AI reliability and accuracy [37][38] Group 9 - The emergence of embodied intelligence is expected to bridge the gap between digital healthcare and physical health services, focusing on autonomous medical interventions and human-robot collaboration [45][49][51] Group 10 - AI technology is fundamentally reshaping the healthcare industry by enhancing operational efficiency and expanding access to quality medical resources [54] - The shift from hospital-centered systems to personal health information systems marks a significant transformation in healthcare delivery, emphasizing proactive health management [56][60]

近日，在国家人工智能应用中试基地（医疗）·浙江举办的行业年度盛会上，"MedAIBench医疗大模型测评榜"正 式揭晓。在这份被誉为"医疗AI国考"的成绩单中，"Ai好医生诊疗支持系统" 同源底层基座模型——全诊医学大模 型（Trizen-Med-Omni）凭借卓越的技术底座，从众多模型中脱颖而出，成功入选"优秀国产医疗大模型"榜单。 2025年，好医生集团与全诊医学携手共进，宣布达成深度战略合作，结合好医生集团的基层网络和全诊医学的人 工智能技术，为全国基层医疗机构提供智能化解决方案，并推出领先的AI大模型"Ai好医生诊疗支持系统"。 目前上线的"Ai好医生诊疗支持系统"，其核心大脑正是此次获奖的Trizen-Med-Omni模型。 系统完美继承了获奖模型的"感知-书写-决策"一体化能力： MedAIBench测评体系由国家人工智能应用中试基地（医疗）联合中国医学科学院北京协和医学院等权威机构打 造，严格遵循国家卫健委《卫生健康行业人工智能应用场景参考指引》。 这一测评结果的公布，实际上确立了2026年医疗大模型的准入"金标准"——即不再仅仅看重参数规模，而是更看 重在智能辅助病历撰写、辅助诊断及医疗安全三 ...

Core Insights - Baichuan Intelligent officially launched the evidence-enhanced medical model Baichuan-M3Plus on January 22, achieving leading results in multiple medical evaluations with a factual hallucination rate reduced to 2.6%, the lowest reported level to date [1] Group 1: Model Features - M3Plus utilizes "evidence anchoring" technology, allowing each medical conclusion generated by the model to be precisely linked to specific evidence paragraphs in original papers or guidelines, ensuring traceability and verifiability of conclusions [1] - The accuracy of matching conclusions to evidence exceeds 95%, indicating a high level of reliability in the model's outputs [1] Group 2: Service Model - Baichuan Intelligent announced the "Haina Baichuan (301667)" plan, which will provide free access to the M3Plus API for Chinese medical service institutions [1] - The API call cost for M3Plus has been reduced by 70% compared to the previous generation, making it more accessible for developers [1] - All developers can apply for a limited-time 15-day free trial of the API, promoting wider adoption of the model [1]

Core Insights - Baichuan Intelligence has launched Baichuan-M3 Plus, which significantly improves accuracy and reliability in serious medical scenarios, surpassing its predecessor M3 and achieving a new world record [2] - The M3 model has outperformed GPT-5.2 in various authoritative evaluations, achieving a hallucination rate of only 3.5%, and M3 Plus further reduces this to 2.6%, a decrease of over 30% compared to GPT-5.2 [2][4] - The introduction of the "Evidence Anchoring" technology ensures that every medical conclusion generated by the model is directly traceable to specific evidence segments in original papers or guidelines, achieving a matching accuracy of over 95% [4] Technical Innovations - M3 Plus employs a unique reinforcement learning paradigm called Fact-Aware RL, which significantly reduces hallucination rates without the use of external tools, achieving state-of-the-art (SOTA) levels [2] - The system architecture has been comprehensively restructured, utilizing key technologies such as MoE architecture optimization and Gated Eagle-3 speculative decoding, resulting in a 70% reduction in API call costs compared to the previous generation [9] Industry Impact - Baichuan has launched the "Haina Baichuan" initiative, offering the world's lowest hallucination evidence-enhanced medical model as a free API to Chinese medical institutions, aiming to foster the domestic AI medical ecosystem [9] - This initiative represents a commitment to advancing AI in healthcare, ensuring that every healthcare worker in China has access to reliable and effective AI assistance, ultimately benefiting public health [9]

新浪科技讯 1月22日上午消息，在"Baichuan-M3 Plus模型发布媒体沟通会"上，百川智能继开源了新一 代医疗大模型Baichuan-M3，时隔9天再次发布Baichuan-M3 Plus模型（简称：M3 Plus），宣布在医疗问 答准确性、可靠性上较M3进一步提升，再次刷新医疗模型低幻觉世界纪录。 据悉，在幻觉率上，M3 Plus相较于此前发布的M3模型大幅下降，在Halluciation Rate评测中由M3的3.5 幻觉率，降低至M3 Plus的2.6。在现场实测环节，百川智能对比了多家国内主流AI医疗模型产品，结果 显示，在多家同行在回答问题引用源、专业表述方式均存在不同程度迷惑性的情况下，接入M3 Plus的 百小应取得了最为准确的回答结果。 此外，M3 Plus还打破了模型成本高的难题，实现API调用降价70%，为规模化应用扫清障碍。 责任编辑：杨赐 新浪科技讯 1月22日上午消息，在"Baichuan-M3 Plus模型发布媒体沟通会"上，百川智能继开源了新一 代医疗大模型Baichuan-M3，时隔9天再次发布Baichuan-M3 Plus模型（简称：M3 Plus），宣布在医疗问 ...

Investment Rating - The report assigns a "Buy" rating for the company, marking its first coverage [10][12]. Core Insights - The company has established a comprehensive GBC (Government, Business, Consumer) business model that integrates AI capabilities across the entire medical service cycle, from health risk warning to chronic disease management [4][20]. - The company is expected to achieve revenues of 920 million, 1.18 billion, and 1.47 billion RMB for the years 2025, 2026, and 2027, respectively, reflecting year-on-year growth rates of 25.6%, 27.6%, and 25.0% [4][10]. Summary by Relevant Sections Company Overview - Founded in 2016, the company leverages the iFlytek Spark Medical Model to provide solutions covering the entire medical service cycle [7][20]. - The ownership structure is concentrated, with iFlytek Group holding 49.4% of shares, ensuring strategic alignment and resource allocation [26][28]. Business Model and Market Position - The company has a significant first-mover advantage, having accumulated extensive data assets through early strategic positioning [8][10]. - The GBC model encompasses a complete medical service loop, addressing challenges in data flow within the healthcare industry [8][10]. Financial Performance - Revenue has shown robust growth, increasing from 373 million RMB in 2021 to an expected 734 million RMB in 2024, with a compound annual growth rate of 25.4% [36]. - The company has improved its net loss from -189 million RMB in 2022 to -133 million RMB in 2024, indicating a positive trend in financial health [37]. Growth Drivers - The company is positioned to benefit from accelerating digitalization and intelligence demands in the healthcare sector, with G and B segments providing stable revenue growth in the short term [10][20]. - The C segment, focusing on patient management services, is anticipated to be a core growth driver in the medium term [10][20]. Technological Advancements - The company has developed the Spark Medical Model, which is the only medical deep reasoning model trained on fully domestic computing power, enhancing its competitive edge [60][61]. - The integration of AI technologies into various healthcare applications is expected to significantly improve operational efficiency and patient outcomes [60][61].