IQuestLab/Fleming-R1-32B
Fleming-R1-32B by IQuestLab is a 32 billion parameter reasoning model built on Qwen3-32B, specifically designed for medical scenarios. It performs step-by-step analysis of complex problems using a "chain-of-thought cold start" and large-scale reinforcement learning. This model excels at medical reasoning, achieving performance comparable to much larger models and strong results on Chinese tasks, with a context length of 32768 tokens.
Loading preview...
Fleming-R1-32B: Medical Reasoning Model
Fleming-R1-32B is a 32 billion parameter language model developed by IQuestLab, fine-tuned on Qwen3-32B, and specialized for medical reasoning tasks. It is engineered to analyze complex medical problems step-by-step, providing reliable answers through a unique training paradigm.
Key Capabilities & Features
- Advanced Medical Reasoning: Designed to perform detailed, step-by-step analysis in medical contexts, addressing complex problems effectively.
- Reinforcement Learning Integration: Utilizes a "chain-of-thought cold start" combined with large-scale reinforcement learning, including adaptive hard-negative mining, to enhance reasoning capabilities, especially for difficult cases.
- Data Strategy: Incorporates public medical datasets with knowledge graphs to broaden coverage of rare diseases, medications, and multi-hop reasoning chains.
- Performance: The 32B version demonstrates performance close to the much larger GPT-OSS-120B and shows superior results on Chinese medical tasks. It surpasses models of similar and larger sizes on the MedXpertQA benchmark.
- Context Length: Supports a substantial context length of 32768 tokens.
Good For
- Medical Scenario Analysis: Ideal for applications requiring detailed reasoning and problem-solving in medical contexts.
- Research and Development: Suitable for researchers and developers exploring advanced AI applications in healthcare, particularly for diagnostic support and treatment planning (with professional oversight).
- Complex Query Resolution: Excels at handling intricate medical queries that require multi-step logical deduction.
Note: This model is intended for research and non-clinical reference only and must not be used for actual diagnosis or treatment decisions without review by qualified professionals. For more details, refer to the research paper.