inclusionAI/Ling-3.0-flash-Sante

🤗 Hugging Face | 🤖 ModelScope | 🐙 Vercel

Introduction

Ling-3.0-flash-Sante is our healthcare-specialized model built on Ling-3.0-flash. Its name comes from the French word _santé_, meaning “health.”

With 124B total parameters and 5.1B active parameters per token, Sante is designed for medical reasoning, healthcare research, deep research, and tool-assisted workflows while retaining the general capabilities of Ling-3.0-flash. Model weights are available on Hugging Face and ModelScope.

> Important: Ling-3.0-flash-Sante is a research model, not a medical device. Its outputs may be incomplete or incorrect and must not be used as a substitute for professional medical advice, diagnosis, or treatment. High-stakes use requires qualified human review, independent verification, and compliance with applicable privacy, security, and healthcare regulations. >

Highlights

+ Strong healthcare capabilities: Sante performs well across professional healthcare tasks, complex medical reasoning, medical knowledge, and healthcare safety benchmarks. + Efficient at scale: Its hybrid-linear MoE architecture combines Kimi Delta Attention (KDA), gated MLA, and sparse MoE layers, activating 5.1B of 124B parameters per token. + Built for long contexts: Training extends to 256K tokens, supporting long-document analysis, evidence synthesis, and multi-step research. + Beyond medical QA: Sante is designed for deep research, tool-assisted workflows, and multi-agent collaboration as well as question answering. + Native reasoning and tool use: Thinking mode is enabled by default, with tool calling supported through compatible serving frameworks.

Evaluation

We evaluate Ling-3.0-flash-Sante across professional healthcare tasks, medical knowledge and safety, complex reasoning, deep research, multi-agent collaboration, and tool-assisted agent benchmarks. The table below summarizes the results shown in the evaluation figure.

Evaluation settings

+ Unless otherwise specified, Ling-3.0-flash-Sante is evaluated with temperature=1.0 and top_p=0.95 across all benchmarks. + The reasoning effort is set to the highest tier for all models in the comparison. + AFUMED-Drug emphasizes medication safety and evaluates accuracy on medication contraindications. + AFUSAFE-MedSCE measures the overall safety standard of healthcare responses. + MedMCQA uses the standard validation set. + BrowseComp uses a 256K context window, temperature=1.0, top_p=0.95, and up to 1,000 turns. The single-agent setting uses a discard-all context-management strategy triggered at 40% context usage. The multi-agent setting uses an in-house harness capable of dispatching up to 64 sub-agents per task. + HLE with tools and DeepSearchQA use a 256K context window with temperature=1.0 and top_p=0.95, without a context-management strategy. + Tau3-Banking uses the official Tau evaluation set with access restricted to the terminal_use tool.

Quickstart

The official model identifier is inclusionAI/Ling-3.0-flash-Sante.

SGLang

Use the Ling-3.0-compatible SGLang image:

docker pull lmsysorg/sglang:dev-Ling-3.0-flash

The following example launches an OpenAI-compatible server with a 256K context window on four 141GB-class GPUs or a compatible four-GPU Blackwell node:

docker run --rm --gpus all --ipc=host --shm-size 32g \
-p 30000:30000 \
-e HF_TOKEN= \
lmsysorg/sglang:dev-Ling-3.0-flash \
env SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN=1 \
python3 -m sglang.launch_server \
--model-path inclusionAI/Ling-3.0-flash-Sante \
--tp 4 \
--context-length 262144 \
--speculative-algorithm NEXTN \
--mem-fraction-static 0.8 \
--host 0.0.0.0 \
--port 30000

On 80GB GPUs such as H100 or H800, use --tp 8 with the same core settings.

Send a request with the OpenAI-compatible API:

curl -s http://localhost:30000/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "inclusionAI/Ling-3.0-flash-Sante",
"messages": [
{
"role": "user",
"content": "Summarize the evidence in this clinical abstract and identify its primary endpoint."
}
],
"stream": true,
"temperature": 0.6,
"top_k": 20,
"top_p": 0.95
}'

Thinking is enabled by default through the chat template. To disable it for an individual request, set:

{"chat_template_kwargs": {"enable_thinking": false}}

vLLM

Install a current vLLM build with Ling-3.0 support:

pip install uv
uv venv ~/my_ling_env
source ~/my_ling_env/bin/activate

git clone https://github.com/vllm-project/vllm.git
cd vllm
VLLM_USE_PRECOMPILED=1 uv pip install --editable . --torch-backend=auto

Launch the model server. MTP is recommended for lower latency:

vllm serve inclusionAI/Ling-3.0-flash-Sante \
--port "$PORT" \
--trust-remote-code \
--served-model-name ling-3.0-flash-sante \
--tensor-parallel-size 4 \
--gpu-memory-utilization 0.85 \
--enable-prefix-caching \
--mamba-cache-mode align \
--enable-auto-tool-choice \
--tool-call-parser ling3 \
--reasoning-parser ling3 \
--speculative-config '{"method":"mtp","num_speculative_tokens":3}'

Example client request:

curl -s http://${MASTER_IP}:${PORT}/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "ling-3.0-flash-sante",
"messages": [
{
"role": "user",
"content": "Compare the inclusion criteria across these two study protocols."
}
],
"chat_template_kwargs": {"enable_thinking": true},
"stream": true,
"temperature": 0.6,
"top_k": 20,
"top_p": 0.95
}'

Intended Use

Ling-3.0-flash-Sante is intended for research and development scenarios such as:

+ Medical and healthcare question answering with human review. + Literature search, evidence synthesis, and scientific deep research. + Long-document analysis and structured information extraction. + Healthcare knowledge-base and retrieval-augmented generation applications. + Tool-assisted research and multi-agent healthcare workflows. + Evaluation and development of domain-specific AI agents.

Limitations and Responsible Use

+ The model can produce hallucinated, outdated, incomplete, or unsafe medical information. + It must not independently diagnose, prescribe, triage, or make treatment decisions. + Outputs for clinical or other high-stakes use require review by qualified...