fauxpaslife/HuatuoGPT-3-9B

fauxpaslife/ HuatuoGPT-3-9B:latest

1 Download Updated 9 hours ago

HuatuoGPT-3-9B is a medical LLM built on Qwen3.5-9B with One-stage Policy Optimization (OnePO). OnePO adapts language models to medicine in a single reinforcement-learning stage, without preceding domain-specific supervised fine-tuning. Q4_K_M Version.

ollama run fauxpaslife/HuatuoGPT-3-9B

curl http://localhost:11434/api/chat \
  -d '{
    "model": "fauxpaslife/HuatuoGPT-3-9B",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='fauxpaslife/HuatuoGPT-3-9B',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'fauxpaslife/HuatuoGPT-3-9B',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 9 hours ago

9 hours ago

a15569ff3ceb · 5.6GB

model

archqwen35

parameters8.95B

quantizationQ4_K_M

5.6GB

template

{{- if .System }}<|im_start|>system {{ .System }}<|im_end|> {{ end -}} {{- range .Messages }}<|im_st

269B

params

{ "stop": [ "<|im_end|>", "<|im_start|>" ], "temperature": 0.2 }

77B

Readme

Good medical candidate, need to test it compared to MedGemma, but functions well thus far. Some of the responses seem choppy.

Note: Lots of thinking, which is important for defining medical accuracy. Also note the speed - could be improved if I were not running other GPU-consuming items. Imported via Forge using standard ChatML/Qwen template. 0.2 temp.

HF Original Model: FreedomIntelligence/HuatuoGPT-3-9B
HF GGUF Model: mradermacher/HuatuoGPT-3-9B-GGUF
Note: This is the Q4_K_M version.

citation:

@inproceedings{chen2026onepo,
  title={OnePO: Direct One-stage Policy Optimization for SFT-free Domain Adaptation},
  author={Chen, Junying and Xie, Xinyuan and Li, Ziniu and Wang, Benyou},
  booktitle={Proceedings of the 43rd International Conference on Machine Learning},
  year={2026}
}