Instructions to use medagent/MedExAgent-8B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use medagent/MedExAgent-8B with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="medagent/MedExAgent-8B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("medagent/MedExAgent-8B")
model = AutoModelForCausalLM.from_pretrained("medagent/MedExAgent-8B", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use medagent/MedExAgent-8B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "medagent/MedExAgent-8B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "medagent/MedExAgent-8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/medagent/MedExAgent-8B

SGLang

How to use medagent/MedExAgent-8B with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "medagent/MedExAgent-8B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "medagent/MedExAgent-8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "medagent/MedExAgent-8B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "medagent/MedExAgent-8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use medagent/MedExAgent-8B with Docker Model Runner:
```
docker model run hf.co/medagent/MedExAgent-8B
```

Research-use acknowledgement required

Access requires agreement to the research-only and no-clinical-use terms below.

MedExAgent-8B is a research model for studying medical-agent behavior in simulated diagnostic conversations. It may ask follow-up questions, request simulated examination or test information, and produce diagnostic reasoning or differential-diagnosis style outputs. It is not a medical device, has not been validated for clinical deployment, and may generate incomplete, incorrect, or unsafe medical information. By requesting access, you agree that you will not use this model for clinical care, diagnosis, treatment decisions, emergency triage, patient management, or any other real-world medical decision-making. You are solely responsible for any research use and for complying with all applicable laws, institutional policies, and safety review requirements.

MedExAgent-8B

Paper: https://arxiv.org/abs/2605.07058

Citation

@misc{gao2026medexagenttrainingllmagents,
      title={MedExAgent: Training LLM Agents to Ask, Examine, and Diagnose in Noisy Clinical Environments}, 
      author={Yicheng Gao and Xiaolin Zhou and Yahan Li and Yue Zhao and Ruishan Liu},
      year={2026},
      eprint={2605.07058},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2605.07058}, 
}

MedExAgent-8B is a research language model for medical-agent experiments. It is intended to study how an AI agent can conduct simulated diagnostic interactions: asking relevant follow-up questions, requesting available simulated examinations or tests, integrating noisy patient information, and producing diagnostic reasoning or differential-diagnosis style responses.

The model is designed for controlled research and evaluation workflows, not for use with real patients. It may be useful for experiments involving simulated patient conversations, diagnostic reasoning benchmarks, medical-agent tool-use studies, and robustness analysis under noisy clinical communication settings.

Intended Use

Use this model only for research, benchmarking, and development in controlled non-clinical environments. Outputs should be treated as unverified model-generated text and independently evaluated by qualified reviewers when relevant.

Clinical Use Disclaimer

This model must not be used for clinical care, diagnosis, treatment decisions, emergency triage, patient management, or any other real-world medical decision-making. It is not a medical device, has not been clinically validated, and may generate incomplete, incorrect, misleading, or unsafe medical information.

By requesting access to this model, you acknowledge and agree that the model is for research use only and that you will not use it in clinical or patient-facing settings.

License

Built with Llama.

MedExAgent-8B is finetuned from Meditron3-8B, which is finetuned from Llama-3.1-8B-Instruct. It is released under the Llama 3.1 Community License and subject to the Llama 3.1 Acceptable Use Policy. Llama 3.1 is licensed under the Llama 3.1 Community License, Copyright © Meta Platforms, Inc. All Rights Reserved.