Instructions to use QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-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 QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-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 QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4
# Run inference directly in the terminal:
llama cli -hf QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4
# Run inference directly in the terminal:
llama cli -hf QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4

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 QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4
# Run inference directly in the terminal:
./llama-cli -hf QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4

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 QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4
# Run inference directly in the terminal:
./build/bin/llama-cli -hf QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4

Use Docker

docker model run hf.co/QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4

LM Studio
Jan

vLLM

How to use QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-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": "QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4

Ollama
How to use QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF with Ollama:
```
ollama run hf.co/QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4
```
Unsloth Desktop

How to use QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4

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": "QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF with Docker Model Runner:
```
docker model run hf.co/QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4
```

Lemonade

How to use QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4

Run and chat with the model

lemonade run user.Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF-NVFP4

List all available models

lemonade list

Hermes Agent

How to use QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-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 QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4

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 QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4

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 "QQZ2026/Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF:NVFP4" \
  --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"

Qwen3.6-27B-NVIDIA-NVFP4-no-MTP-GGUF / ATTRIBUTION.md

QQZ2026

Publish model card, license, and attribution

a671b16 verified 2 months ago

preview code

Raw

History Blame Contribute Delete

1.16 kB

	# Attribution and modification notice

	## Upstream works

	This derivative is based on the following upstream works:

	1. Qwen/Qwen3.6-27B — base model by the Qwen team.
	2. nvidia/Qwen3.6-27B-NVFP4 — NVIDIA ModelOpt NVFP4 checkpoint.
	3. utautako/Qwen3.6-27B-NVIDIA-NVFP4-MTP-GGUF — GGUF conversion preserving NVFP4 tensors and including one embedded MTP layer.
	4. ggml-org/llama.cpp — GGUF tooling and inference runtime.

	Each upstream model/repository should be consulted for its own model card, acceptable-use guidance, license metadata, limitations, and acknowledgements.

	## Modification

	Wilson Zhang created a derivative GGUF by:

	- pruning layer `64`, which is the embedded MTP / next-token-prediction layer;
	- overriding `qwen35.block_count` from `65` to `64`;
	- overriding `qwen35.nextn_predict_layers` from `1` to `0`;
	- copying the remaining tensors without requantizing them;
	- validating full-GPU loading at 65,536 context on an RTX 5060 Ti 16GB with Q4_0 K/V cache.

	This modification does not claim authorship of the base model, NVIDIA checkpoint, upstream GGUF conversion, quantization method, or llama.cpp implementation.