Instructions to use wredd/medgemma-4b-gguf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use wredd/medgemma-4b-gguf with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="wredd/medgemma-4b-gguf",
	filename="medgemma-4b-iq1_m.gguf",
)

llm.create_chat_completion(
	messages = "No input example has been defined for this model task."
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use wredd/medgemma-4b-gguf with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf wredd/medgemma-4b-gguf:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf wredd/medgemma-4b-gguf:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf wredd/medgemma-4b-gguf:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf wredd/medgemma-4b-gguf:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf wredd/medgemma-4b-gguf:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf wredd/medgemma-4b-gguf:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf wredd/medgemma-4b-gguf:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf wredd/medgemma-4b-gguf:Q4_K_M

Use Docker

docker model run hf.co/wredd/medgemma-4b-gguf:Q4_K_M

LM Studio
Jan
Ollama
How to use wredd/medgemma-4b-gguf with Ollama:
```
ollama run hf.co/wredd/medgemma-4b-gguf:Q4_K_M
```

Unsloth Studio

How to use wredd/medgemma-4b-gguf with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for wredd/medgemma-4b-gguf to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for wredd/medgemma-4b-gguf to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for wredd/medgemma-4b-gguf to start chatting

Atomic Chat new
Docker Model Runner
How to use wredd/medgemma-4b-gguf with Docker Model Runner:
```
docker model run hf.co/wredd/medgemma-4b-gguf:Q4_K_M
```

Lemonade

How to use wredd/medgemma-4b-gguf with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull wredd/medgemma-4b-gguf:Q4_K_M

Run and chat with the model

lemonade run user.medgemma-4b-gguf-Q4_K_M

List all available models

lemonade list

MedGemma 4B GGUF - Quantized for African Healthcare

Quantized versions of google/medgemma-1.5-4b-it optimized for on-device medical AI in resource-constrained settings.

Available Models

File	Quantization	Size	RAM	MedQA (vs Unquantized)	Use Case
`medgemma-4b-iq2_xs.gguf`	IQ2_XS (2-bit) + Medical imatrix	~0.85GB	~1.9GB	43.8% (63% retained)	Ultra-budget phones
`medgemma-4b-q2_k.gguf`	Q2_K (2-bit)	~1.3GB	~2.3GB	34.7% (50% retained)	Budget phones
`medgemma-4b-q4_k_m.gguf`	Q4_K_M (4-bit)	~2.3GB	~3GB	56.0% (81% retained)	Budget phones

Medical Importance Matrix

The IQ2_XS model was quantized using a custom importance matrix (imatrix) calibrated on:

African primary care scenarios (malaria, typhoid, cholera, respiratory infections)
Maternal and child health (pregnancy complications, childhood diarrhea, nutrition)
Emergency triage (snake bites, severe dehydration, trauma)
Multi-language symptoms (Twi, Hausa, Yoruba, English)

This preserves medical diagnostic accuracy while aggressively compressing general knowledge.

Usage with llama.cpp

./llama-cli -m medgemma-4b-iq2_xs.gguf -p "Patient has fever, chills, and headache for 3 days. What could this be?"

License

Subject to Gemma Terms of Use.

Part of the Nku Project

Built for the Google MedGemma Impact Challenge - bringing AI-powered healthcare to underserved African communities.

Downloads last month: 57

GGUF

Model size

4B params

Architecture

gemma3

Hardware compatibility

1-bit

2-bit

4-bit

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for wredd/medgemma-4b-gguf

Base model

google/medgemma-1.5-4b-it

Quantized

(34)

this model