Instructions to use pfnet/Preferred-MedLLM-Qwen-72B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use pfnet/Preferred-MedLLM-Qwen-72B with Transformers:

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

pipe = pipeline("text-generation", model="pfnet/Preferred-MedLLM-Qwen-72B")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("pfnet/Preferred-MedLLM-Qwen-72B", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use pfnet/Preferred-MedLLM-Qwen-72B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "pfnet/Preferred-MedLLM-Qwen-72B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "pfnet/Preferred-MedLLM-Qwen-72B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/pfnet/Preferred-MedLLM-Qwen-72B

SGLang

How to use pfnet/Preferred-MedLLM-Qwen-72B 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 "pfnet/Preferred-MedLLM-Qwen-72B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "pfnet/Preferred-MedLLM-Qwen-72B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "pfnet/Preferred-MedLLM-Qwen-72B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "pfnet/Preferred-MedLLM-Qwen-72B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use pfnet/Preferred-MedLLM-Qwen-72B with Docker Model Runner:
```
docker model run hf.co/pfnet/Preferred-MedLLM-Qwen-72B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Preferred-MedLLM-Qwen-72B

Model Description

Preferred-MedLLM-Qwen-72B is a finetuned model based on Qwen/Qwen2.5-72B, which has undergone continued pretraining on an original corpus of medical-related text.

The model is released under the Qwen LICENSE.

Model Performance

The table below shows the performance on the Japanese medical licensing examination from 2018 to 2022 (IgakuQA).

Model ID	Average	2018	2019	2020	2021	2022
Preferred-MedLLM-Qwen-72B	431.2	434	420	439	430	433
GPT-4o	430.4	427	431	433	427	434
Qwen2.5-72B	398.4	412	394	394	393	399
Llama3-Preferred-MedSwallow-70B	395.2	407	390	391	393	395
GPT-4	388.8	382	385	387	398	392
Mistral-Large-Instruct-2407	376	370	371	390	373	376
Llama-3.1-Swallow-70B-v0.1	368.4	379	378	379	351	355
Meta-Llama-3-70B	334.6	353	340	348	314	318
GPT-3.5	273.2	266	250	266	297	287

Limitations

The model was developed for research purposes and is not intended for clinical diagnosis. It is the users' responsibility to ensure compliance with applicable rules and regulations.

Contributors

Preferred Networks, Inc.

Junichiro Iwasawa
Wataru Kawakami
Keita Suzuki

Publications

Detailed evaluation results are given in the blog and research paper.

Citations

@article{preferredmedllm2025,
      title={Stabilizing Reasoning in Medical LLMs with Continued Pretraining and Reasoning Preference Optimization}, 
      author={Kawakami, Wataru and Suzuki, Keita and Iwasawa, Junichiro},
      journal={arXiv preprint arXiv:2504.18080},
      year={2025}
}