Instructions to use DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-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 DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-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 DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-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/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-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/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-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/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF:Q4_K_M

Use Docker

docker model run hf.co/DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-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": "DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-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/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF:Q4_K_M

Ollama
How to use DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF with Ollama:
```
ollama run hf.co/DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF:Q4_K_M
```
Unsloth Desktop

How to use DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-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 DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF:Q4_K_M

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": "DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF with Docker Model Runner:
```
docker model run hf.co/DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF:Q4_K_M
```

Lemonade

How to use DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-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 DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-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/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-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 DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-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/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-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"

[BUG] INSTRUCT mode stops suddenly: does not call tools in multi-turn agent loops

by BebopVox - opened 14 days ago

Discussion

BebopVox

14 days ago

Hi,
Thank you for your model! 💟

I have a problem with INSTRUCT mode.

The problem below reproduces 11/Sep/2026 by myself. Investigation and text below is written with help of AI

Summary

The model ends its turn after a narrative intent sentence. It does not emit the promised tool call. This breaks multi-turn agentic tool loops. Enabling low-effort thinking fixes the problem.

Environment

Item	Value
Llama-server	build b10641, Vulkan backend
Docker image	`image: kyuz0/amd-strix-halo-toolboxes:vulkan-radv-performance`
Hardware	AMD Strix Halo, 64 GB unified memory
Quant	`Qwen3.8-27B-TTURBO-Fable-C-Fusion-709-L-Uncen-NM-DAU-NEO-Q4_K_M.gguf`
Harness	OpenAI-compatible chat completions with tools / or OMP (oh-my-pi) agent

Server flags:

--parallel 2
--temp 0.7
--top-p 0.80
--top-k 20
--presence-penalty 0.0
--repeat-last-n 0
--repeat-penalty 1.0

Chat-template-kwargs:

{"enable_thinking":false, "reasoning_effort":"xhigh", "preserve_thinking":true}

Observed symptom

The model calls the first tool correctly. The tool result returns. The model then writes a short sentence, for example:

"Let me also list files to demonstrate another tool:"

The turn ends. It emits no second tool call. Logs show stopReason:"stop", hasToolCalls:false, hasText:true.

The user must prompt again before the model continues. The model repeats this pattern when prompted to continue.

Reproduction

Start a chat with tools available.
Ask the model to demonstrate tool use.
Let it call the first tool.
Return one tool result.
Observe the next turn.

Expected result: the model calls the next tool.

Actual result: the model narrates intent and stops.

Measured chain rate

Same two-turn reproduction run repeatedly:

Thinking mode	Chained second tool	Rate
Off (instruct, effort xhigh)	1 of 12	~8%
Low	7 of 8	88%
Medium	8 of 8	100%

The failure is stochastic. The same configuration flips between success and failure between runs.

Root cause hypothesis

The TURBO fusion compresses reasoning tokens. In instruct mode there is no reasoning block at all. The model appears to plan the next tool call internally. The output generation then stops before it can emit the <tool_call>. This matches the issue reported on the TURBO-735 card in discussions #22 and #25.

Workaround that works

Set enable_thinking:true and reasoning_effort:low. The model then chains tools reliably at a modest reasoning-token cost.

Questions

Is this expected or anyone faces this as well on INSTRUCT mode (not thinking)?
Is reasoning_effort:low the recommended mode for agentic tool loops on TWIN-TURBO?

mythrime

14 days ago

https://huggingface.co/peculiar-ragdoll/Qwen-Sharp-Chat-Templates

Try using this as your chat template

DavidAU

Owner 14 days ago

@BebopVox

thank you for your report and detailed testing/notes.

Try quant:
Q5KS/Q6 ; as tool calling (with instruct mode) will need the extra precision... you might even need Q8.

Hopefully this should help.

BebopVox

14 days ago

Tried both recommendations.
Still no success with Qwen3.8-27B-TTURBO-Fable-C-Fusion-709-L-Uncen-NM-DAU-NEO-Q6_K.gguf 😥

I'll try Q8 later on.

BebopVox changed discussion title from [Bug] INSTRUCT mode stops after narration: does not call tools in multi-turn agent loops to [BUG] INSTRUCT mode stops suddenly: does not call tools in multi-turn agent loops 14 days ago

BebopVox

14 days ago

Q8 same problem
Qwen3.8-27B-TTURBO-Fable-C-Fusion-709-L-Uncen-NM-DAU-NEO-Q8_0.gguf

BebopVox

14 days ago

Do you have any ideas what am I doing wrong?
Or have you reproduced the same behavior?

I wanted to use instruct to make this model x2-x3 faster and remove thinking whatsoever.
Yes this model is cooking but dense 27B is still very slow on my hardware so Im looking for alternatives 😊

bioleyl

14 days ago

I have the same problem. Also tried multiple quantization and settings but it seems to be related to context length. As soon as I reach ~20K context it's a non stop failing. Never before.

shahariarrabby

13 days ago

I am also having exact same problem with llamacpp + opencode

DavidAU

Owner 13 days ago

FIX and new jinja (with tool call issue resolved) located here:

https://huggingface.co/DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF/discussions/10

DavidAU changed discussion status to closed 11 days ago

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