Model Details

license: apache-2.0 language: - uk - en base_model: google/gemma-3-4b-it library_name: gguf pipeline_tag: text-generation tags: - medical - cardiology - gemma-3 - quantized

Gemma-3-4B-Amosov-v1-GGUF

Model Overview

This repository contains the GGUF quantized version of Gemma-3-4B-Amosov-v1, a specialized medical LLM based on the Google Gemma-3-4B-IT architecture.

The model is specifically fine-tuned to automate cardiological recommendations and clinical decision support according to the protocols . It has been trained on a specialized dataset of 12,000+ medical records and epicrises, focusing on Ischemic Heart Disease (IHD) and Atrial Fibrillation.

Quantization Details

File: gemma-amosov-Q4_K_M.gguf
Method: Q4_K_M (Llama.cpp k-quants)
Original Precision: BF16
Description: 4-bit quantization that offers a significant reduction in model size and memory usage while maintaining high accuracy and reasoning capabilities.

Architecture Highlights

Base Model: Gemma-3-4B-IT
Context Length: 131,072 tokens (Max Position Embeddings)
Model Type: gemma3 (Gemma3ForConditionalGeneration)

1. Requirements

To run this model, you need llama.cpp or any GGUF-compatible software (LM Studio, Jan, Ollama).

2. Recommended Inference Settings

Based on the fine-tuning process, the following parameters are recommended for medical accuracy:

Temperature: 0.01 (near-deterministic)
Do Sample: False
Repetition Penalty: 1.3
Top-P: 0.9

3. Prompt Template (Gemma 3)

The model expects the standard Gemma 3 chat format. If a system prompt is used, it should be prepended to the first user message:

<start_of_turn>user
{system_prompt}

{user_query}<end_of_turn>
<start_of_turn>model

Example System Prompt:

"You are a cardiologist. Provide only clear medical recommendations based on clinical data. Do not include administrative or non-medical information."

Medical Disclaimer This model is a research tool and is intended for informational purposes only. It is not a licensed medical device. All outputs, medication dosages, or treatment plans generated by the model must be reviewed and verified by a qualified healthcare professional. Do not use this model for self-diagnosis or self-treatment.

Credits Base Model: google/gemma-3-4b-it

4. Limitations

Optimized for cardiology and cardiac surgery
Reduced accuracy outside these domains
No vision capabilities (text-only MedGemma IT)
May generate incomplete or generalized recommendations

5. Citing & Authors

If you use this model in your research, please cite:

@misc{Ostashko2025MedGemmaCardiology, title = {MedGemma-4B-Cardiology: A Domain-Finetuned Clinical LLM for Cardiology}, author = {Uaritm}, year = {2025}, url = {ai.esemi.org} }

Project homepage: https://ai.esemi.org

LicenseThe use of this model is subject to the terms of the original Gemma License. Please review and adhere to the associated licensing terms for the base model.

Downloads last month: 52

GGUF

Model size

4B params

Architecture

gemma3

Hardware compatibility

4-bit