turkerberkdonmez/TUSGPT-TR-Medical-9B

TEXT GENERATIONConcurrency Cost:1Model Size:9BQuant:FP8Ctx Length:16kPublished:Feb 11, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

TUSGPT-TR-Medical-9B by turkerberkdonmez is a 9-billion parameter language model based on the Gemma-2 architecture, specifically fine-tuned for Turkish medical applications. It was trained using over 55,000 high-quality Turkish medical question-answer pairs across various disciplines, employing a 2-Stage DoRA methodology. This model is designed to advance medical AI research in Turkey by providing accurate and comprehensive responses to medical queries in Turkish.

Loading preview...

TUSGPT-TR-Medical-9B: Turkish Medical Language Model

TUSGPT-TR-Medical-9B, developed by turkerberkdonmez, is Turkey's first open-source Turkish medical language model. Built upon the Gemma-2 architecture with 9 billion parameters, it specializes in medical AI research and applications within the Turkish language.

Key Capabilities & Training

  • Specialized Fine-tuning: The model is fine-tuned on the ytu-ce-cosmos/Turkish-Gemma-9b-T1 base model.
  • Extensive Medical Dataset: It leverages 55,465 high-quality Turkish medical question-answer pairs from the turkerberkdonmez/TUSGPT-TR-Medical-Dataset-v1 dataset.
  • Broad Medical Coverage: The training data spans numerous medical disciplines, including Basic Sciences (Pharmacology, Pathology, Anatomy, Physiology) and Clinical Sciences (Internal Medicine, Surgery, Pediatrics, Obstetrics & Gynecology), as well as Emergency Medicine, Neurology, Oncology, and Radiology.
  • Advanced Training Methodology: Utilizes a 2-Stage DoRA (Weight-Decomposed Low-Rank Adaptation) method, trained on Apple Silicon hardware with bfloat16 precision.

Usage & Recommendations

  • Python (Transformers): Provides clear examples for integration, recommending temperature=0.6, top_p=0.95, top_k=20, and min_p=0 for generation.
  • GGUF Quantized Versions: Available in Q8_0 (9.8 GB) and Q4_K_M (5.8 GB) for local deployment with Ollama and LM Studio, suitable for various hardware configurations.

Important Disclaimer

This model is intended for educational and research purposes only and should not be used as the sole basis for clinical decision-making. Users are advised to consult qualified physicians for medical decisions.

Finetunes
1 models