Instructions to use amkkk/Trace-Inverter-4B-NoBubble-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 amkkk/Trace-Inverter-4B-NoBubble-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 amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0
# Run inference directly in the terminal:
llama cli -hf amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0
# Run inference directly in the terminal:
llama cli -hf amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0

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 amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0
# Run inference directly in the terminal:
./llama-cli -hf amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0

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 amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0
# Run inference directly in the terminal:
./build/bin/llama-cli -hf amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0

Use Docker

docker model run hf.co/amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0

LM Studio
Jan

vLLM

How to use amkkk/Trace-Inverter-4B-NoBubble-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "amkkk/Trace-Inverter-4B-NoBubble-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": "amkkk/Trace-Inverter-4B-NoBubble-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0

Ollama
How to use amkkk/Trace-Inverter-4B-NoBubble-GGUF with Ollama:
```
ollama run hf.co/amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0
```
Unsloth Desktop

How to use amkkk/Trace-Inverter-4B-NoBubble-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0

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": "amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use amkkk/Trace-Inverter-4B-NoBubble-GGUF with Docker Model Runner:
```
docker model run hf.co/amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0
```

Lemonade

How to use amkkk/Trace-Inverter-4B-NoBubble-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0

Run and chat with the model

lemonade run user.Trace-Inverter-4B-NoBubble-GGUF-Q8_0

List all available models

lemonade list

Hermes Agent

How to use amkkk/Trace-Inverter-4B-NoBubble-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 amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0

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 amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use amkkk/Trace-Inverter-4B-NoBubble-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0

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 "amkkk/Trace-Inverter-4B-NoBubble-GGUF:Q8_0" \
  --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"

Trace-Inverter-4B-NoBubble GGUF

This is the GGUF Q8_0 release of amkkk/Trace-Inverter-4B-NoBubble. It contains the same merged Trace-Inverter-4B-NoBubble weights converted for llama.cpp-compatible runtimes.

Trace-Inverter-4B-NoBubble is a 4B-parameter no-bubble trace inversion model. Given an original problem or conversation context and a known final answer, it reconstructs a detailed synthetic reasoning trace wrapped in <think> and </think>. No reasoning bubble, reasoning summary, scratchpad, or hidden chain-of-thought is required as input.

Generated traces are synthetic reconstructions. They must not be presented as the actual hidden reasoning of Claude, Qwen, or any source model.

File

Trace-Inverter-4B-NoBubble-Q8_0.gguf - GGUF Q8_0 quantization

The GGUF preserves the model tokenizer and chat template metadata from the BF16 Transformers release.

Base Model

Qwen/Qwen3-4B-Instruct-2507

Relationship To BF16 Release

This GGUF is a quantized release of the same weights hosted at amkkk/Trace-Inverter-4B-NoBubble. See the BF16 model card for full training details, dataset provenance, limitations, and benchmark methodology.

This is the NoBubble inverter. It is not Jackrong/Trace-Inverter-4B and does not consume reasoning bubbles at inference time.

Usage

With llama.cpp:

llama-cli -m Trace-Inverter-4B-NoBubble-Q8_0.gguf -p "<your chat-formatted prompt>" -n 4096

With Ollama, create a Modelfile next to the GGUF:

FROM Trace-Inverter-4B-NoBubble-Q8_0.gguf
PARAMETER temperature 0
PARAMETER num_predict 4096

Then run:

ollama create trace-inverter-4b-nobubble-gguf -f Modelfile
ollama run trace-inverter-4b-nobubble-gguf

Recommended prompt shape:

Problem:
{problem}

Model's final answer:
{final_answer}

Reconstruct the detailed synthetic reasoning trace.

Expected output:

<think>
...synthetic reconstructed reasoning...
</think>

Public-10 GGUF Check

Deterministic public-10 evaluation was run locally through llama.cpp server against data/processed/public10.jsonl.

Metric	BF16 published	GGUF Q8_0
Token F1	0.6500	0.6567
ROUGE-L	0.3916	0.3981
Length ratio	0.9366	0.9568
Format pass	100.0%	100.0%
`<tool_call>` rate	0.0%	0.0%

License

Apache 2.0. The BF16 release, base model, and source datasets are Apache 2.0.

Downloads last month: -

GGUF

Model size

4B params

Architecture

qwen3

Hardware compatibility

8-bit

Model tree for amkkk/Trace-Inverter-4B-NoBubble-GGUF

Base model

Qwen/Qwen3-4B-Instruct-2507

Quantized

(298)

this model