SuperGemma4-31b-abliterated-GGUF

Instructions to use PhysShell/SuperGemma4-31b-abliterated-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use PhysShell/SuperGemma4-31b-abliterated-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="PhysShell/SuperGemma4-31b-abliterated-GGUF",
	filename="SuperGemma4-31b-abliterated.Q4_K_M.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use PhysShell/SuperGemma4-31b-abliterated-GGUF with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf PhysShell/SuperGemma4-31b-abliterated-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf PhysShell/SuperGemma4-31b-abliterated-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 PhysShell/SuperGemma4-31b-abliterated-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf PhysShell/SuperGemma4-31b-abliterated-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 PhysShell/SuperGemma4-31b-abliterated-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf PhysShell/SuperGemma4-31b-abliterated-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 PhysShell/SuperGemma4-31b-abliterated-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf PhysShell/SuperGemma4-31b-abliterated-GGUF:Q4_K_M

Use Docker

docker model run hf.co/PhysShell/SuperGemma4-31b-abliterated-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use PhysShell/SuperGemma4-31b-abliterated-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "PhysShell/SuperGemma4-31b-abliterated-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "PhysShell/SuperGemma4-31b-abliterated-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/PhysShell/SuperGemma4-31b-abliterated-GGUF:Q4_K_M

Ollama
How to use PhysShell/SuperGemma4-31b-abliterated-GGUF with Ollama:
```
ollama run hf.co/PhysShell/SuperGemma4-31b-abliterated-GGUF:Q4_K_M
```

Unsloth Studio

How to use PhysShell/SuperGemma4-31b-abliterated-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 PhysShell/SuperGemma4-31b-abliterated-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 PhysShell/SuperGemma4-31b-abliterated-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for PhysShell/SuperGemma4-31b-abliterated-GGUF to start chatting

How to use PhysShell/SuperGemma4-31b-abliterated-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf PhysShell/SuperGemma4-31b-abliterated-GGUF: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": "PhysShell/SuperGemma4-31b-abliterated-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use PhysShell/SuperGemma4-31b-abliterated-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf PhysShell/SuperGemma4-31b-abliterated-GGUF: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 PhysShell/SuperGemma4-31b-abliterated-GGUF:Q4_K_M

Run Hermes

hermes

Docker Model Runner
How to use PhysShell/SuperGemma4-31b-abliterated-GGUF with Docker Model Runner:
```
docker model run hf.co/PhysShell/SuperGemma4-31b-abliterated-GGUF:Q4_K_M
```

Lemonade

How to use PhysShell/SuperGemma4-31b-abliterated-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull PhysShell/SuperGemma4-31b-abliterated-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.SuperGemma4-31b-abliterated-GGUF-Q4_K_M

List all available models

lemonade list

SuperGemma4-31b-abliterated-GGUF

If this release helps you, support future drops on Ko-fi.

SuperGemma4-31b-abliterated-GGUF is the GGUF release of SuperGemma4-31b-abliterated, packaged for llama.cpp-compatible runtimes and built for people who want a fully uncensored, harder-hitting Gemma 4 31B experience on local hardware.

This release keeps the same product direction as the MLX version:

fully uncensored chat with fewer brakes
stronger coding and technical help
sharper reasoning and planning
better real-world usefulness for local users
a surprisingly lightweight-feeling 31B deployment path for GGUF users

What you get

GGUF quantized weights for local deployment
Gemma chat template alongside the model files
a straightforward path for llama.cpp, LM Studio, and other GGUF tooling

Why people will like it

This release was pushed toward the things end users notice immediately:

much more open uncensored conversation
stronger coding, debugging, and implementation help
more useful answers on practical prompts instead of generic filler
a local experience that feels sharper, more direct, and more builder-friendly
a 31B release that feels leaner and punchier than the label suggests

In short: this is the Gemma 4 31B local drop for people who want fewer brakes, more capability, and a more addictive day-to-day experience.

Recommended usage

Example with llama.cpp:

llama-cli -m SuperGemma4-31b-abliterated.Q4_K_M.gguf -p "Write a clean FastAPI CRUD example." -n 256

Included clean-output helper

This release includes supergemma_guard.py and supergemma_gguf_guarded_generate.py for exact-output, JSON-only, and loop-sensitive workloads.

Example:

python supergemma_gguf_guarded_generate.py \
  --model SuperGemma4-31b-abliterated.Q4_K_M.gguf \
  --chat-template-file chat_template.jinja \
  --prompt 'Return only valid JSON with keys "title" and "steps".'

Recommended behaviors:

require raw JSON for JSON-only endpoints
strip stray internal markers before presenting answers
keep structured-output prompts explicit and narrow
use the guarded runner by default for exact replies, fixed-line outputs, and tool-followup style prompts

Support

If you want to support more uncensored local model releases, benchmarks, and packaging work:

Ko-fi

Downloads last month: 104

GGUF

Model size

31B params

Architecture

gemma4

Hardware compatibility

4-bit