Instructions to use ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-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 ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-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 ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S
# Run inference directly in the terminal:
llama cli -hf ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S
# Run inference directly in the terminal:
llama cli -hf ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S

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 ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S
# Run inference directly in the terminal:
./llama-cli -hf ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S

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 ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S
# Run inference directly in the terminal:
./build/bin/llama-cli -hf ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S

Use Docker

docker model run hf.co/ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S

LM Studio
Jan

vLLM

How to use ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-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": "ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-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/ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S

Ollama
How to use ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF with Ollama:
```
ollama run hf.co/ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S
```
Unsloth Desktop

How to use ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S

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": "ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF with Docker Model Runner:
```
docker model run hf.co/ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S
```

Lemonade

How to use ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S

Run and chat with the model

lemonade run user.Qwen3.8-27B-GSQ-RCO-GGUF-IQ2_S

List all available models

lemonade list

Hermes Agent

How to use ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-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 ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S

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 ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S

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 "ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF:IQ2_S" \
  --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"

Mtp integrated version coming?

by CheesingGouda - opened 13 days ago

Discussion

CheesingGouda

13 days ago

Is there a version planned or in the works with mtp integrated?

cruizba

12 days ago

I was able to run it with MTP by merging the MTP of the Unsloth UD_IQ3_XXS model with this one! But yes it would be cool to have official MTPs, it will be probably faster.

It is an amazing model for what I used it.

cr404

12 days ago

+1 this is a very promising model and it could be even better with the MTP intact.

@cruizba how did you merge these?

CheesingGouda

12 days ago

•

edited 12 days ago

Yeah, i see a ud q4 mtp model is all? I'm currently using jrell/Qwen3.8-27B-i1-IQ4_XS-GGUF-Smaller and I'm hopeful that this new model, once it gets mtp will give me even more gpu VRAM space!

cruizba

12 days ago

•

edited 12 days ago

@cr404 @CheesingGouda I've published my merge of Unsloth MTPs in a HF repo: https://huggingface.co/cruizba/ISTA-DASLab-Qwen3.8-27B-GSQ-RCO-GGUF-Unsloth-MTP

The repo also includes the graft script if you want to make your own.

How I merged it: the GSQ file ships without the 15 blk.64.* nextn tensors, so --spec-type draft-mtp can't run on it. The head is native to Qwen3.8-27B and unsloth's UD quants keep it, so I copied it over with a small gguf-py script. Three changes: copy the 15 tensors, set qwen35.nextn_predict_layers = 1, and bump qwen35.block_count to 65: llama.cpp places the head at block_count - nextn, so leaving it at 64 fails at load. I grafted from the UD quant rather than the standalone MTP sidecar because the head shares the target's embedding/output (~195 MiB vs ~1 GB duplicated).

Quality is untouched: weights and perplexity are byte/last-digit identical to the originals, and at temp > 0 llama.cpp only keeps a drafted token if it equals what the model would sample.

On my 16 GB card (IQ3_XXS): greedy 45 → 93 tok/s @160K, +32% e2e under real sampling (temp 1.0 / top-p 0.95 / top-k 20). Caveat: the draft costs ~1 GB of VRAM, so you trade some context for speed (160K vs 216K on IQ3_XXS).

NOTE: This is completely unofficial and experimental, in fact it is my first experiment uploading a model 😅 .
I hope ISTA-DASLab ships official MTP GGUFs 😃

anm2211

IST Austria Distributed Algorithms and Systems Lab org 12 days ago

Thank you everyone for participating in this discussion and for all the valuable feedback and insights!

We also wanted to share a quick update: MTP models and higher-bitwidth models will be released tomorrow.

Thanks again for your interest, testing, and contributions!

cruizba

12 days ago

•

edited 11 days ago

@anm2211 Thank you for your work!!

forceware680

12 days ago

Yess im waiting for official MTP support. thanks for hardwork

helcig

IST Austria Distributed Algorithms and Systems Lab org 11 days ago

@cruizba @forceware680 We just published the MTP version, thank you for you feedback!

Andelys

10 days ago

•

edited 10 days ago

Where are they , i do not find them ?
Qwen3.8-27B-GSQ-RCO-mtp_iqX._.gguf not found the one i found are the qunatification Qwen3.8-27B-GSQ-RCO-IQx_x-mtp.gguf of 10to 12.1GB but no draft-mtp
Looking forward to test in full with high curiosity
best R

anm2211

IST Austria Distributed Algorithms and Systems Lab org 10 days ago

The files ending in -mtp.gguf are the MTP-integrated versions.

For example, you can use Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp.gguf and enable it with:
--spec-type draft-mtp

So you don’t need a separate mtp_iqX...gguf draft file. The MTP head is included directly in the *-mtp.gguf model.

Andelys

10 days ago

•

edited 10 days ago

ok thx it was llama.cpp with an uncorrect buld , i have different version with different patch , so fixed it , works very very beautifully . thx

Definitely Looking forward for IQ4 , as IQ3 works like a charm....very good job

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