Instructions to use Mediform/gemma4-v5 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Mediform/gemma4-v5 with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="Mediform/gemma4-v5",
	filename="gemma4-e4b-v5-Q4_K_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 Mediform/gemma4-v5 with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Mediform/gemma4-v5:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf Mediform/gemma4-v5:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Mediform/gemma4-v5:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf Mediform/gemma4-v5: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 Mediform/gemma4-v5:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf Mediform/gemma4-v5: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 Mediform/gemma4-v5:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Mediform/gemma4-v5:Q4_K_M

Use Docker

docker model run hf.co/Mediform/gemma4-v5:Q4_K_M

LM Studio
Jan
Ollama
How to use Mediform/gemma4-v5 with Ollama:
```
ollama run hf.co/Mediform/gemma4-v5:Q4_K_M
```

Unsloth Studio

How to use Mediform/gemma4-v5 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 Mediform/gemma4-v5 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 Mediform/gemma4-v5 to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for Mediform/gemma4-v5 to start chatting

How to use Mediform/gemma4-v5 with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf Mediform/gemma4-v5:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "Mediform/gemma4-v5:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use Mediform/gemma4-v5 with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf Mediform/gemma4-v5:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default Mediform/gemma4-v5:Q4_K_M

Run Hermes

hermes

Atomic Chat new
Docker Model Runner
How to use Mediform/gemma4-v5 with Docker Model Runner:
```
docker model run hf.co/Mediform/gemma4-v5:Q4_K_M
```

Lemonade

How to use Mediform/gemma4-v5 with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Mediform/gemma4-v5:Q4_K_M

Run and chat with the model

lemonade run user.gemma4-v5-Q4_K_M

List all available models

lemonade list

gemma-4-E4B-it — Scribion German Medical Extraction v5 (recall-focused) — GGUF Q4_K_M

LoRA fine-tune of google/gemma-4-E4B-it for German consultation fact extraction in Scribion's exact 5-call schema ({id,text,section,evidence} over the 24 NOTE_STATE paths). v5 is the recall-focused successor to v3 (Mediform/gemma4): it keeps v3's section-routing fixes but recovers facts v3 omitted.

Files

gemma4-e4b-v5-Q4_K_M.gguf — language model, Q4_K_M (~5.3 GB).
mmproj-gemma-4-E4B-it-Q8_0.gguf — stock multimodal projector (audio+vision), unchanged.

v5 vs v3 (held-out Scribion clips, max_tokens 4096)

clip	v3	v5
arztbericht	18/22 facts	20/22 (recovers Vacoped orthosis, 2 crutches, Clexane, Fersensporn, knee finding)
froehlich	12/17	12/17

What changed in training: the per-type empty calls were rebalanced 52%→20% (removing the sparsity bias that made v3 under-extract) and the teacher targets were regenerated with an exhaustiveness hint (capture negatives, devices, already-given meds, minor facts — no hallucination). Routing/dedup discipline from v3 is preserved (Weber-B → diagnosen_gesichert/ befunde, not vitalparameter).

Trade-off: higher recall at a small precision cost — slightly more verbose, occasional cross-section duplicate (e.g. a fact placed in two plausibly-valid sections). Net: better for the importance-weighted Q metric (which penalizes omissions far more than minor dups). Validate on your DeepSeek Q pipeline before promoting over v3.

Usage (llama.cpp)

llama-cli -m gemma4-e4b-v5-Q4_K_M.gguf -sys "<scribion per-type system prompt>" -p "Transkript:\n..."

Note: use maxTokens≈4096 for the narrative call — exhaustive output can be long; a smaller cap silently truncates the JSON (→ omitted facts). Decode low-temp (≈0.05) for deterministic extraction.

Quantized with llama.cpp (CPU build).

Downloads last month: -

GGUF

Model size

7B params

Architecture

gemma4

Hardware compatibility

4-bit

Inference Providers NEW

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

Model tree for Mediform/gemma4-v5

Base model

google/gemma-4-E4B

Finetuned

google/gemma-4-E4B-it

Quantized

(233)

this model