Instructions to use DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF",
	filename="Qwen3.6-27B-NEO-CODE-HERE-2T-OT-HIGH-Q8_0.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": [
				{
					"type": "text",
					"text": "Describe this image in one sentence."
				},
				{
					"type": "image_url",
					"image_url": {
						"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
					}
				}
			]
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-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 DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-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 DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-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 DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-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 DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF:Q4_K_M

Use Docker

docker model run hf.co/DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-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": "DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF:Q4_K_M

Ollama
How to use DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF with Ollama:
```
ollama run hf.co/DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF:Q4_K_M
```

Unsloth Studio

How to use DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-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 DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-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 DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF to start chatting

How to use DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-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": "DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-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 DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-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 DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-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 "DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-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 DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF with Docker Model Runner:
```
docker model run hf.co/DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF:Q4_K_M
```

Lemonade

How to use DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF-Q4_K_M

List all available models

lemonade list

MTP Speculative Decoding Version

by apollo-mg - opened May 13

Discussion

apollo-mg

May 13

Would you consider producing a version with MTP support?

Thanks,
Mark

DavidAU

Owner May 14

Not at this time, as it is unclear if heretic/transformers/unsloth support it fully.
Each must fully support it ; then the model must be "re-heretic'ed", re-trained, and re-quanted/tested/benchmarked.

kamjin

May 14

I've already seen a heretic version based on Qwen 3.6 27B on Huggingface that supports MTP, model address: https://huggingface.co/llmfan46/Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved-GGUF, but I'm unsure how it's done and what the actual results are.

willworker

May 17

•

edited May 17

For anyone looking to just bolt on the Q8.0 MTP layers to the this model (regardless of quant), you can currently find the MTP weights and a convert.py that will bolt them on to this (or any Qwen3.6 27B model) here: https://huggingface.co/havenoammo/Qwen3.6-27B-MTP-UD-GGUF

If that repo disappears, just ping me. I have the weights and script backed up.

For clarity, I can only confirm that this works for making this model load with MTP and go zoom-zoom with the code. I have no idea what it does to the benchmarks. However, since the main model is supposed to reject any drafts it would not itself produce, there shouldn't be any impact other increased speed.

I have been seeing about an 80% to 100% increase TPS , but a 50% decrease in TTFT. This appears to be expected and known behavour with the MTP implementation in llama.cpp, which has taken a "stable deployment first, optimize later" stance on getting it out. I anticipate there will be refinements over the coming weeks.

GreenWoods

May 19

I'm really looking forward to adding MTP on this model.This model is so wonderful.

ElvisM

May 20

Hoping for this release, too. Qwen models are powerful, but sadly, not very useful for creative purposes. We can only rely on finetunes, but unfortunately, they don't have MTP.

ElvisM

May 21

Holy, just tested it here with the regular Qwen 27b. 20 tokens/s on a 4060ti +2060 (22gb VRAM total) is a game changer. I don't care much about the lower PP speed. This finally makes the model usable for me. Hopefully this becomes the norm for all future models.

DavidAU

Owner May 22

Still some time before MTP version.

As of this writing:

There are pipeline (issues as well as optimizations) issues still currently, and it is not widely supported in some AI Apps.

Specifically:

Ggufs:

Imatrix is not yet supported for MTP.
Not all AI apps have updated to support it -> result -> MTP ggufs do not work at all.
Misc issues with speed still being worked on.

Training:

Unknown if support has been merged into Transformers (previously stripped out/ignored MTP tensors)
Unsloth and BNB support. (unknown status)

As soon as the "pipeline" settles...

JamesNot007

May 24

Remind me when MTB version comes.

ai-henry

May 30

Hi DavidAU, just wanted to give you props - this model has been absolutely amazing in my opinion even in comparison to using enterprise frontiers during the day. Really hoping you'll release an MTP version. Thanks!

DavidAU

Owner May 31

@ai-henry Thanks so much.

shokac

8 days ago

any update on MTP version ?

DavidAU

Owner 6 days ago

Awaiting some pipeline issue - MTP Imatrix, LLamacpp; and in the process of tuning system upgrades atm.

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