Model release · Ant Group
Ling-3.0-flash-Sante
- Released
- Review updated
A healthcare-enhanced MoE model for medical research and workflows, with improvements in deep research, multi-agent collaboration and tool use. Ant Ling reports that HealthBench Professional rose from 45.73 to 54.29, ranking first among flash-size models in its evaluation. AFUMED-ComplexReasoning rose from 65.22 to 76.92, while MedQA-USMLE scored 94.21. Reported BrowseComp scores are 80.2 for a single agent and 90.47 for a multi-agent setup; HLE with tools scored 57.11.
Availability
Model weights and expanded evaluation results are available.