HuatuoGPT is an open-source large language model trained on a vast Chinese medical corpus to behave like a knowledgeable, patient-friendly doctor. Hosted on GitHub with about 1,300 stars under the Apache-2.0 license, the project from FreedomIntelligence was accepted at EMNLP 2023 and its code and weights were updated through 2024.
The goal of HuatuoGPT is a more professional ChatGPT-style assistant for medical consultation scenarios. It is trained with a hybrid supervised fine-tuning dataset that combines distilled ChatGPT responses with real doctor-patient data, aiming to give the model both doctor-like expertise and a patient-friendly tone, followed by reinforcement learning from AI feedback to improve its answers. The repository releases the HuatuoGPT-7B and 13B model weights, the SFT data, evaluation data and a separately released reward model, along with an online demo. A successor, HuatuoGPT-II, has since pushed results further on Chinese medical benchmarks, but this original repository remains a widely referenced starting point for open medical LLMs. Because the training data recipe, evaluation data and reward model are all public, follow-up projects can inspect exactly how the model was built and reproduce its results.
HuatuoGPT is free and open source under the Apache-2.0 license. The code, data and model weights are available from GitHub and Hugging Face at no cost; you only supply the hardware to run or fine-tune the models.
Researchers and developers building Chinese medical AI โ consultation assistants, medical question answering and health information tools โ who need an open, fine-tunable foundation model with a medical focus. It is a research model, not a substitute for professional medical advice.