OpenMeditron/Meditron3-Qwen2.5-14B: A Medical LLM for Research
OpenMeditron/Meditron3-Qwen2.5-14B is a 14.8 billion parameter large language model (LLM) developed by the OpenMeditron initiative, specifically specialized in clinical medicine. It is built upon the Qwen2.5-14B base model and is co-designed with expert clinicians and humanitarian practitioners.
Key Characteristics & Specialization
- Medical Focus: The model's training emphasizes equitable representation, contextual diversity, and actionable real-world evidence-based guidelines within clinical medicine.
- Global Context: Particular effort has been made to represent limited-resource and humanitarian settings, neglected populations, and diseases.
- Training Data: Trained on a diverse mixture of expert-curated public data, including clinical guidelines, peer-reviewed medical publications, synthetic differential diagnoses, replay data, and LLM-enhanced medical MCQs.
- Research-Only: Meditron-3 is intended solely for research to study and evaluate the potential of LLMs in clinical decision-making and accessing evidence-based medical information. It is not validated for direct medical use.
Performance & Downstream Use
While primarily a research model, Meditron-3 shows competitive performance on medical multiple-choice question benchmarks. For instance, it achieves an average of 71.09% across MedmcQA, MedQA, and PubmedQA, slightly outperforming the base Qwen/Qwen2.5-14B-Instruct model. The model is a foundation model that can be adapted for specific downstream tasks like question-answering through in-context learning or fine-tuning.
Important Note
This model is a static release trained on an offline dataset. It is provided "AS IS" and is not intended for clinical decision-making, diagnosis, or treatment of patients. Users should consult qualified healthcare providers.