inclusionAI releases Ling-3.0-flash-Sante, an open healthcare research model

inclusionAI has released Ling-3.0-flash-Sante, an open-weight model built on its Ling-3.0-flash base and specialized for healthcare. The name comes from the French word for health, and the model has 124 billion total parameters with 5.1 billion active per token, using a hybrid architecture that mixes Kimi Delta Attention, gated MLA and sparse mixture-of-experts layers. Training extends to a 256K-token context window, which inclusionAI says supports long-document analysis and multi-step research.

inclusionAI says the model is aimed at medical reasoning, healthcare research, deep research and tool-assisted, multi-agent workflows, with thinking mode and tool calling enabled by default, and that it performs well on the company's own professional healthcare, medical knowledge and safety benchmarks. The weights are posted on Hugging Face and ModelScope, with setup instructions for serving through SGLang or vLLM.

inclusionAI is explicit that Ling-3.0-flash-Sante is a research model, not a medical device. The company warns its outputs can be hallucinated, outdated, incomplete or unsafe, that it must not be used to independently diagnose, prescribe or triage, and that any clinical or high-stakes use requires review by qualified professionals and verification against authoritative sources.

Why it matters

It gives developers an open-weight model to build and test healthcare AI tools with, as long as outputs stay behind human review.

Source: inclusionAI on Hugging Face