FreedomIntelligence/HuatuoGPT-3-27B · Hugging Face

FreedomIntelligence has released HuatuoGPT-3-27B, a medical large language model built on Qwen3.8-27B. This model utilizes One-stage Policy Optimization (OnePO), a reinforcement learning method that adapts language models to medical applications in a single stage, bypassing the need for prior domain-specific supervised fine-tuning. Teacher responses offer initial guidance and are phased out as the model's performance improves. This follows their previous release of HuatuoGPT-3-9B last week.