Ant open-sources Ling-3.0-flash-Sante medical AI model
The benchmark mix matters: it beats larger frontier models on diagnosis-style multiple-choice tasks while staying slightly behind on expert-level text reasoning, so it is a specialist rather than a general replacement. Developers and medical researchers can self-deploy and fine-tune it for literature research, clinical guideline retrieval, drug fact-checking and evidence curation apps. As an open-weight 124B MoE model, it lowers the cost barrier for healthcare AI teams in ASEAN and Chinese-speaking markets.
Ant Group released Ling-3.0-flash-Sante, an open-source medical-enhanced model built on Ling-3.0-flash. It uses a hybrid attention and sparse MoE architecture with 124B total parameters, 5.1B activated, and a 256K context window. The model supports thinking mode, tool calling and multi-agent collaboration for tasks such as literature retrieval, comparing clinical study designs and analysing long medical documents. On DiagnosisArena-MCQ it scores 83.8, above Kimi K3's 78.4 and GPT-5.6 Sol's 81.9; it reaches 86.9 on BrowseComp in multi-agent configuration and 53.9 on MedXpertQA-Text.
蚂蚁刚刚把其医疗增强模型给开源了：Ling-3.0-flash-Sante，重点面向专业医疗任务和复杂医学推理
让医疗AI能检索医学文献、进行复杂医学推理、调用工具，用多个Agent协作完成医疗研究任务
基于Ling-3.0-flash的强化，混合注意力、稀疏MoE架构，总参数124B，激活5.1B，256K上下文
面向医疗研究的Agent能力，支持思考模式、工具调用和多Agent协作，可以围绕医学问题检索文献、比较临床研究方案、整理证据、分析长篇医疗文档等
在BrowseComp上，多Agent配置下达到了86.9； DiagnosisArena-MCQ上83.8，高于Kimi K3的78.4、GPT-5.6 Sol的81.9； MedXpertQA-Text 达到53.9，略高于Kimi K3 53.5，低于 GPT-5.6 Sol 60.2、Gemini 3.6 Flash 62.4
支持自主部署，搞医疗的可以拿去研究或二开。拿来构建医学文献研究、临床指南检索、药物知识核查、科研证据整理应用等
#LingflashSante #AI医疗模型
