Instructions to use yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-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 yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-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 yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-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 yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-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 yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-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 yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF:Q4_K_M

Use Docker

docker model run hf.co/yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-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": "yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF:Q4_K_M

Ollama
How to use yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF with Ollama:
```
ollama run hf.co/yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF:Q4_K_M
```

Unsloth Studio

How to use yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-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 yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-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 yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF to start chatting

How to use yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-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": "yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-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 yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-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 yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-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 "yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-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 yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF with Docker Model Runner:
```
docker model run hf.co/yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF:Q4_K_M
```

Lemonade

How to use yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.gemma-4-12B-coder-fable5-composer2.5-v1-GGUF-Q4_K_M

List all available models

lemonade list

Thank you and there is interest!

by NullSense - opened Jun 14

Discussion

NullSense

Jun 14

Thank you for this upload this is amazing! Would love a v2 with more fable data. If there's a way to contribute more data, please do add instructions!

yuxinlu1

Owner Jun 14

Thanks, really glad it's useful! 🙏 The tricky part is that Fable 5 access got pulled, so I can't generate new Fable CoT data anymore — what I have is what I saved beforehand. So v2 will lean more on the verifiable Composer data as the backbone.
But yeah, contributions are very welcome! The most useful format is verifiable coding CoT — basically: a problem, a full chain-of-thought solution, and test cases that actually pass when run. That way I can filter by whether the tests pass and keep only clean samples. If you've got data like that (especially for languages/domains that are underrepresented), open an issue or ping me and I'll put together proper contribution instructions. 👍

OblxDev

Jun 14

Wether it's with fable or any decent model, i'll definitely try it too, looks amazing in current state already !

NullSense

Jun 17

•

edited Jun 17

Thanks, really glad it's useful! 🙏 The tricky part is that Fable 5 access got pulled, so I can't generate new Fable CoT data anymore — what I have is what I saved beforehand. So v2 will lean more on the verifiable Composer data as the backbone.
But yeah, contributions are very welcome! The most useful format is verifiable coding CoT — basically: a problem, a full chain-of-thought solution, and test cases that actually pass when run. That way I can filter by whether the tests pass and keep only clean samples. If you've got data like that (especially for languages/domains that are underrepresented), open an issue or ping me and I'll put together proper contribution instructions. 👍

I do have some fable 5 and a lot of opus 4.8 traces, but fable 5 is like half a day of coding really, not sure if that's enough... Rust, Python, TS.
Question though: are you planning on doing any benchmarks? It's a bit unclear to me still if any of these fine tunes from traces of thicker models actually result in better coding/reasoning ability.

yuxinlu1

Owner Jun 18

@NullSense Both would be great — and half a day of Fable 5 is plenty, even a few clean samples help.

The Opus 4.8 traces are exactly what I want, and Rust/TS are underrepresented so especially useful. Ideal format is problem + full
CoT + runnable tests, so I can auto-keep only the ones whose tests actually pass; no tests is fine for the rarer languages, I'll
just spot-check those. For the Fable 5 ones I have to verify provenance before I can label anything "Fable 5" — so please include
the raw session export with the model ID. Open an issue or ping me and I'll send a small template.

On benchmarks — fair to be skeptical, and it's really the key question. For v1 all I have so far is a small held-out test — real but tiny. For v2 I'm planning to run proper benchmarks —
Tau2 (agentic) for sure, and most likely LiveCodeBench (coding) — to actually see whether distilling
traces from a thicker model moves coding/reasoning. Honest expectation: coding and agentic gain the most; raw knowledge is the
hardest to shift, since traces don't add facts.

NullSense

Jun 19

@NullSense Both would be great — and half a day of Fable 5 is plenty, even a few clean samples help.

The Opus 4.8 traces are exactly what I want, and Rust/TS are underrepresented so especially useful. Ideal format is problem + full
CoT + runnable tests, so I can auto-keep only the ones whose tests actually pass; no tests is fine for the rarer languages, I'll
just spot-check those. For the Fable 5 ones I have to verify provenance before I can label anything "Fable 5" — so please include
the raw session export with the model ID. Open an issue or ping me and I'll send a small template.

On benchmarks — fair to be skeptical, and it's really the key question. For v1 all I have so far is a small held-out test — real but tiny. For v2 I'm planning to run proper benchmarks —
Tau2 (agentic) for sure, and most likely LiveCodeBench (coding) — to actually see whether distilling
traces from a thicker model moves coding/reasoning. Honest expectation: coding and agentic gain the most; raw knowledge is the
hardest to shift, since traces don't add facts.

That sounds great! I'd be down to get you traces I run opus 4.8 high all day on multiple repos - python, ts, rust and on infra stuff, would love to help, just really not sure how to package all this. Curious about these finetunes - does turning off thinking mode affect things? i.e. are the fine tunes specifically affecting thinking mode?

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