Instructions to use neko-legends/Qwen3.6-27B-NVFP4-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 neko-legends/Qwen3.6-27B-NVFP4-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 neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF:NVFP4
# Run inference directly in the terminal:
llama cli -hf neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF:NVFP4

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF:NVFP4
# Run inference directly in the terminal:
llama cli -hf neko-legends/Qwen3.6-27B-NVFP4-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 neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF:NVFP4
# Run inference directly in the terminal:
./llama-cli -hf neko-legends/Qwen3.6-27B-NVFP4-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 neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF:NVFP4
# Run inference directly in the terminal:
./build/bin/llama-cli -hf neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF:NVFP4

Use Docker

docker model run hf.co/neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF:NVFP4

LM Studio
Jan

vLLM

How to use neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "neko-legends/Qwen3.6-27B-NVFP4-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": "neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF:NVFP4

Ollama
How to use neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF with Ollama:
```
ollama run hf.co/neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF:NVFP4
```
Unsloth Desktop

How to use neko-legends/Qwen3.6-27B-NVFP4-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 neko-legends/Qwen3.6-27B-NVFP4-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": "neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF:NVFP4"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF with Docker Model Runner:
```
docker model run hf.co/neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF:NVFP4
```

Lemonade

How to use neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF:NVFP4

Run and chat with the model

lemonade run user.Qwen3.6-27B-NVFP4-MTP-GGUF-NVFP4

List all available models

lemonade list

Hermes Agent

How to use neko-legends/Qwen3.6-27B-NVFP4-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 neko-legends/Qwen3.6-27B-NVFP4-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 neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF:NVFP4

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use neko-legends/Qwen3.6-27B-NVFP4-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 neko-legends/Qwen3.6-27B-NVFP4-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 "neko-legends/Qwen3.6-27B-NVFP4-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"

llama.cpp doesn't detect the model as capable of reasoning

by Gaboo - opened Jul 20

Discussion

Gaboo

Jul 20

Latest llama.cpp on Win 11 with the powershell command:

llama serve `
    -hf "neko-legends/Qwen3.6-27B-NVFP4-MTP-GGUF:NVFP4_UNSLOTH" `
    --port 1234 `
    --jinja `
    --reasoning on `
    --reasoning-preserve

The model loads but does not reason and all responses are given instantly. I tried the chat template as well which also did not work either --chat-template-kwargs '{"enable_thinking":true,"preserve_thinking":true}'. The llama.cpp web interface also does not label it as capable of reasoning. Oddly, llama.cpp recognizes that this model is capable of preserving reasoning, but not reasoning.

neko-legends

Owner Jul 20

Thanks for reporting this. What client or harness are you using to send requests after starting llama.cpp: the built-in web UI, Open WebUI, SillyTavern, Continue, or something else?

I tested the exact NVFP4_UNSLOTH GGUF on Windows 11 using llama.cpp b10069 and called /v1/chat/completions directly. With --jinja --reasoning on --reasoning-preserve, the response contained a full choices[0].message.reasoning_content field followed by the normal final answer.

Test prompt: “A farmer has 17 sheep. All but 9 run away. Think carefully and explain how many remain.”
The model generated roughly 460 reasoning tokens before answering. As a control, I sent the same request with chat_template_kwargs.enable_thinking=false; it returned an immediate 47-token answer with reasoning_content: null, which matches the behavior you described.

This suggests the GGUF and embedded template are working, but the client/harness may not detect reasoning support or may be overriding the server with enable_thinking:false. Also, llama.cpp’s /v1/models currently lists this as completion only, so that capability list is not a reliable reasoning indicator.

Could you check your browser’s Network tab or share the JSON request body being sent to /v1/chat/completions? In particular, look for chat_template_kwargs.enable_thinking.

Gaboo

Jul 20

I’m using llama.cpp b10069 with CUDA 13.3. Previously I was testing through the built-in web UI rather than calling the API directly.

I tested /v1/chat/completions and reasoning works correctly there. With chat_template_kwargs.enable_thinking=true the response included a populated reasoning_content field, and with it set to false, reasoning_content was null.

So this appears to be an issue with the built-in web harness. Specifically, it seems that when the web harness fails to detect reasoning capability in the model, it forcefully disables reasoning and ignores any presets that enabled it.

As shown in the image above it forcefully sets enable_thinking to false, presumably because it could not detect the models capabilities.

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