Instructions to use SciTools/gemma with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use SciTools/gemma with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf SciTools/gemma:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf SciTools/gemma:Q4_K_M

Install from WinGet (Windows)

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

Use Docker

docker model run hf.co/SciTools/gemma:Q4_K_M

LM Studio
Jan
Ollama
How to use SciTools/gemma with Ollama:
```
ollama run hf.co/SciTools/gemma:Q4_K_M
```
Unsloth Desktop

How to use SciTools/gemma with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf SciTools/gemma:Q4_K_M

Configure the model in Pi

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

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use SciTools/gemma with Docker Model Runner:
```
docker model run hf.co/SciTools/gemma:Q4_K_M
```

Lemonade

How to use SciTools/gemma with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull SciTools/gemma:Q4_K_M

Run and chat with the model

lemonade run user.gemma-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use SciTools/gemma with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf SciTools/gemma: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 SciTools/gemma:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use SciTools/gemma with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf SciTools/gemma:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "SciTools/gemma:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Gemma 4 model weights in GGUF format, used by Understand's AI features and served locally by ullama (llama.cpp).

gemma-4-E4B-it-qat-UD-Q4_K_XL.gguf — what Understand's Gemma4-E4B choice downloads. Google's quantization-aware-trained (QAT) E4B as quantized by Unsloth, copied unmodified from unsloth/gemma-4-E4B-it-qat-GGUF (SHA-256 df0fd4ee07072c607c29a0a1cb4f98918426cca12f45a2776bdd6ee6d09a4de3). Tested against the file below on an Apple M5 MacBook Pro and an RTX 4090 Linux machine, it qualified on both. It followed instructions better (0.944 vs 0.897 on the M5, 0.906 vs 0.844 on Linux), answered sooner (51 s vs 62 s median on the M5) and wrote more accurate code summaries (0.663 vs 0.639 on the M5), in a 4.2 GB download instead of 5.0 GB. Its chat answers matched the code about as closely (0.694 vs 0.721 on the M5, 0.749 vs 0.727 on Linux).
gemma-4-E4B-it-Q4_K_M.gguf — the earlier E4B choice, qualified in SciTools' two-workload evaluation (chat quality 148.6 on the frozen GitAhead scale, 3/3 qualifying runs). Kept for installs that already use it.
gemma-4-12b-it-Q4_K_M.gguf — what Understand's Gemma4-12B choice downloads, copied unmodified from unsloth/gemma-4-12b-it-GGUF (SHA-256 0a270ec9fe6b34f4a0d33992b6135117b484ebc4766ab76b51d4ae8c457e4c42). Tested on an Apple M5 MacBook Pro, it qualified and read the code most thoroughly of the gemma models before answering (0.903), but answers slowest of them, in a 7.1 GB download.

Ship and use exactly these files: verdicts do not carry across quantizations.

Gemma 4 is released by Google under the Apache 2.0 license. This repository redistributes the weights unmodified apart from quantization; it is not endorsed by Google.

Downloads last month: 36

GGUF

Model size

12B params

Architecture

gemma4

Hardware compatibility

4-bit

Inference Providers NEW

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