Instructions to use ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF 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 ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-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 ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-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 ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF:Q4_K_M

Use Docker

docker model run hf.co/ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF:Q4_K_M

LM Studio
Jan
Ollama
How to use ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF with Ollama:
```
ollama run hf.co/ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF:Q4_K_M
```

Unsloth Studio

How to use ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-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 ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-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 ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF to start chatting

How to use ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-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": "ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-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 ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF: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 "ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF: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"

Docker Model Runner
How to use ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF with Docker Model Runner:
```
docker model run hf.co/ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF:Q4_K_M
```

Lemonade

How to use ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull ManniX-ITA/gemma-4-A4B-98e-v6-coder-it-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.gemma-4-A4B-98e-v6-coder-it-GGUF-Q4_K_M

List all available models

lemonade list

Are there any plans for the qwen3.6-35B model?

by jian2023 - opened May 26

Discussion

jian2023

May 26

I hope similar trimming will be done on the Qwen 3.6-35B model as well. It would be perfect if these models could have MTP added.

ManniX-ITA

Owner May 26

If I have enough time I will try making a run on Qwen 3.6-35B A3B; the problem is that they are strong on coding but weak on formatting and instructions following.
Pruning can be really destructive on these aspects and Gemma-4 holds it well because it's exceptionally strong there.
I doubt I can get somewhere close to Gemma-4 A4B going down to 20b or 26b.

Google did not release MTP layers for Gemma-4 A4B yet and they would need to be pruned as well.
I've been promised a 2 weeks unlimited testing on Linode, if that materializes then I will try creating my MTP layers for 98e from scratch.
Will need 8TB storage and 2-4 x H100 only for that for a few days, it's uncharted territory for me.

ManniX-ITA

Owner May 26

Oh well I missed that the drafter was released 3 weeks ago, will work on that.

JasonMMo

Jun 12

Thank's You! I'm trying to run gemma on my laptop using llama.ccp/TurboQuant Had a lot of issues but most of them resolved.
So, Basically. I had fetched the regular and then the unsloth and noticed that even the gemini api where all doing the same "Loopdumb".
I had gemini in the browser fight me and help in a lot of ways one way is it brought me to the Template?/Page? Angry, Mage!

And i noticed you had posted about the "SO OBVIOUS THAT YOU COULD NOT IGNORE IT AT ANY STEP "Cough:Google" ISSUES EVEN WITH ALL SYSTEMS CHECKED"
So i clicked on it looked and said "This guys gunna screw it up for them folks at that there Alphabet.inc" Got you fix! Thank you for pointing this out. Saw your
models and was absolutly tempted so i grabbed the average Q_4_K_S v5 and then noticed on ollama it was replaced by v6 so i have that one instead.
To make a long story amount to absolutely nothing, Waste time power and induce a state of semi psyco anticipation. ...................................................................................................

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment