FreedomIntelligence/HuatuoGPT-3-9B
HuatuoGPT-3-9B by FreedomIntelligence is a 9 billion parameter medical large language model built on Qwen3.5-9B, specifically adapted for medical reasoning. It utilizes a novel One-stage Policy Optimization (OnePO) method, directly applying reinforcement learning without prior domain-specific supervised fine-tuning. This model is designed to provide medical insights and requires a 'thinking mode' for inference, generating reasoning before its final answer.
Loading preview...
HuatuoGPT-3-9B: Medical Reasoning LLM
HuatuoGPT-3-9B, developed by FreedomIntelligence, is a 9 billion parameter medical large language model based on the Qwen3.5-9B architecture. Its core innovation lies in the One-stage Policy Optimization (OnePO) method, which directly adapts the language model to the medical domain using reinforcement learning. This approach bypasses traditional supervised fine-tuning stages, making it a distinct method for domain adaptation.
Key Capabilities & Features
- Specialized Medical Reasoning: Designed specifically for complex medical queries and diagnostic support.
- OnePO Training: Leverages a unique reinforcement learning strategy for domain adaptation, guided by teacher responses that are phased out as the model improves.
- Thinking Mode Inference: Requires
enable_thinking=Trueduring inference, allowing the model to generate internal reasoning steps before producing its final medical answer. - Open-source Resources: FreedomIntelligence provides the training code, a medical RL dataset, and an 8B rubric grader for further research and development.
When to Use This Model
- Medical Question Answering: Ideal for applications requiring detailed and reasoned responses to medical inquiries.
- Research in Medical AI: Useful for researchers exploring novel domain adaptation techniques, particularly OnePO.
- Clinical Decision Support Systems: Can serve as a component in systems that require robust medical reasoning capabilities.