Typhoon-Si-Med-Thinking-4B
Domain-Specific (Medical)
Qwen 3
4B
First-of-its-kind ranked-list medical reasoning model (Research Preview)
Released
January 9, 2025
Context
Input
Text
Output
Text
About this Model
First-of-its-kind ranked-list medical reasoning model jointly developed by Typhoon (SCB 10X) and the Siriraj Informatics and Data Innovation Center (SiData+) at Siriraj Hospital, Mahidol University. This 4-billion-parameter instructive model is trained with reinforcement learning to generate ranked lists of candidate answers, giving users breadth and multiple perspectives. Despite its lightweight footprint, it performs robustly across multiple formats—multiple choice, short answer, and ranked list reasoning.
Key Features
Ranked-List Medical Reasoning
Generates ordered lists of plausible diagnoses or answers instead of a single prediction, reflecting real-world clinical reasoning and uncertainty.
Clinically Aligned Answer Formats
Produces structured outputs designed to mirror how clinicians think, supporting differential diagnosis and comparative evaluation.
Reinforcement-Tuned Ranking Accuracy
Fine-tuned with reinforcement learning to prioritize correct answers higher in the list using MRR-style rewards and semantic evaluation.
Small Model, State-of-the-Art Performance
Achieves leading performance on medical reasoning benchmarks with only 4B parameters, outperforming larger frontier models in ranked-list tasks.
Release History
Version 1
January 9, 2025
Initial release
Availability
Web Playground
Not available
Typhoon API
Not available
Typhoon API Pro
Not available
Other Platforms
Not available
Related Links