Instructions to use unsloth/Qwen3.6-35B-A3B-MTP-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use unsloth/Qwen3.6-35B-A3B-MTP-GGUF with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="unsloth/Qwen3.6-35B-A3B-MTP-GGUF")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("unsloth/Qwen3.6-35B-A3B-MTP-GGUF", device_map="auto")

llama-cpp-python

How to use unsloth/Qwen3.6-35B-A3B-MTP-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="unsloth/Qwen3.6-35B-A3B-MTP-GGUF",
	filename="BF16/Qwen3.6-35B-A3B-BF16-00001-of-00002.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 unsloth/Qwen3.6-35B-A3B-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 unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-Q4_K_M
# Run inference directly in the terminal:
llama cli -hf unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-Q4_K_M
# Run inference directly in the terminal:
llama cli -hf unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-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 unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-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 unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-Q4_K_M

Use Docker

docker model run hf.co/unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-Q4_K_M

LM Studio
Jan

vLLM

How to use unsloth/Qwen3.6-35B-A3B-MTP-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "unsloth/Qwen3.6-35B-A3B-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": "unsloth/Qwen3.6-35B-A3B-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/unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-Q4_K_M

SGLang

How to use unsloth/Qwen3.6-35B-A3B-MTP-GGUF with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "unsloth/Qwen3.6-35B-A3B-MTP-GGUF" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "unsloth/Qwen3.6-35B-A3B-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 images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "unsloth/Qwen3.6-35B-A3B-MTP-GGUF" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "unsloth/Qwen3.6-35B-A3B-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"
						}
					}
				]
			}
		]
	}'

Ollama
How to use unsloth/Qwen3.6-35B-A3B-MTP-GGUF with Ollama:
```
ollama run hf.co/unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-Q4_K_M
```

Unsloth Studio

How to use unsloth/Qwen3.6-35B-A3B-MTP-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 unsloth/Qwen3.6-35B-A3B-MTP-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 unsloth/Qwen3.6-35B-A3B-MTP-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for unsloth/Qwen3.6-35B-A3B-MTP-GGUF to start chatting

How to use unsloth/Qwen3.6-35B-A3B-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 unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-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": "unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use unsloth/Qwen3.6-35B-A3B-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 unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-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 unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-Q4_K_M

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use unsloth/Qwen3.6-35B-A3B-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 unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-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 "unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-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 unsloth/Qwen3.6-35B-A3B-MTP-GGUF with Docker Model Runner:
```
docker model run hf.co/unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-Q4_K_M
```

Lemonade

How to use unsloth/Qwen3.6-35B-A3B-MTP-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-Q4_K_M

Run and chat with the model

lemonade run user.Qwen3.6-35B-A3B-MTP-GGUF-UD-Q4_K_M

List all available models

lemonade list

Stable MTP first release!

by danielhanchen - opened May 12

Discussion

danielhanchen

Unsloth AI org May 12

•

edited May 13

MTP GGUFs are still experimental, but for now they function ok

MTP speculative decoding for ~1.5-2x faster generation — build llama.cpp from the MTP PR branch

Thanks for waiting - all quants should work well now - but remember these are still EXPERIMENTAL until the MTP branch is merged

apt-get install pciutils build-essential cmake curl libcurl4-openssl-dev -y
git clone -b mtp-clean https://github.com/am17an/llama.cpp.git
cmake llama.cpp -B llama.cpp/build -DBUILD_SHARED_LIBS=OFF -DGGML_CUDA=ON
cmake --build llama.cpp/build --config Release -j --clean-first --target llama-cli llama-server
cp llama.cpp/build/bin/llama-* llama.cpp

export LLAMA_CACHE="unsloth/Qwen3.6-35B-A3B-MTP-GGUF"
./llama.cpp/llama-server \
    -hf unsloth/Qwen3.6-35B-A3B-MTP-GGUF:UD-Q4_K_XL \
    -ngl 99 -c 8192 -fa on -np 1 \
    --spec-type mtp --spec-draft-n-max 2

Set -DGGML_CUDA=OFF for CPU/Metal. -np > 1 and --mmproj are not yet supported with MTP.

danielhanchen pinned discussion May 12

DaBaiTu

May 12

This comment has been hidden (marked as Resolved)

sedative7626

May 12

•

edited May 12

I have 2t/s on this model and 25t/s on stable llama.cpp without MTP. I don't know why. I compiled with -DGGML_CUDA=OFF. Both models are MXFP4_MOE. Mac M1 Pro 32GB.
My command:

./llama-server --host 127.0.0.1 --port 5678 --flash-attn on --jinja -c 4096 -ngl all --fit on -np 1 -ctk q8_0 -ctv q8_0 --spec-type mtp --spec-draft-n-max 3 -m ~/LLM/models/unsloth/Qwen3.6/Qwen3.6-35B-A3B-MXFP4_MOE-2.gguf

Anyone has the same issue? Did i missed some flag?

slot update_slots: id  0 | task 0 | prompt processing done, n_tokens = 12, batch.n_tokens = 4
~llama_io_write_device: allocated 'MTL0' buffer 62.812 MiB
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5

Is MTP uses much more memory?

DodoPapa

May 12

I have 2t/s on this model and 25t/s on stable llama.cpp without MTP. I don't know why. I compiled with -DGGML_CUDA=OFF. Both models are MXFP4_MOE. Mac M1 Pro 32GB.
My command:

./llama-server --host 127.0.0.1 --port 5678 --flash-attn on --jinja -c 4096 -ngl all --fit on -np 1 -ctk q8_0 -ctv q8_0 --spec-type mtp --spec-draft-n-max 3 -m ~/LLM/models/unsloth/Qwen3.6/Qwen3.6-35B-A3B-MXFP4_MOE-2.gguf

Anyone has the same issue? Did i missed some flag?

slot update_slots: id  0 | task 0 | prompt processing done, n_tokens = 12, batch.n_tokens = 4
~llama_io_write_device: allocated 'MTL0' buffer 62.812 MiB
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5

Is MTP uses much more memory?

yes

danielhanchen

Unsloth AI org May 13

Try reducing --spec-draft-n-max 3 to --spec-draft-n-max 2

sedative7626

May 13

Try reducing --spec-draft-n-max 3 to --spec-draft-n-max 2

Still 2t/s and crashes. Are there are any recommendations how to calculate VRAM required to run this MTP models? E.g., if i previously was running MXFP4 with success, how many VRAM i needed now? x2? x3?

tatdatnguyen21

May 13

Try reducing --spec-draft-n-max 3 to --spec-draft-n-max 2

Still 2t/s and crashes. Are there are any recommendations how to calculate VRAM required to run this MTP models? E.g., if i previously was running MXFP4 with success, how many VRAM i needed now? x2? x3?

Try adding --fitt 2048. If it works, try reducing it. if possible.

LadyJun

May 13

I have 2t/s on this model and 25t/s on stable llama.cpp without MTP. I don't know why. I compiled with -DGGML_CUDA=OFF. Both models are MXFP4_MOE. Mac M1 Pro 32GB.
My command:

./llama-server --host 127.0.0.1 --port 5678 --flash-attn on --jinja -c 4096 -ngl all --fit on -np 1 -ctk q8_0 -ctv q8_0 --spec-type mtp --spec-draft-n-max 3 -m ~/LLM/models/unsloth/Qwen3.6/Qwen3.6-35B-A3B-MXFP4_MOE-2.gguf

Anyone has the same issue? Did i missed some flag?

slot update_slots: id  0 | task 0 | prompt processing done, n_tokens = 12, batch.n_tokens = 4
~llama_io_write_device: allocated 'MTL0' buffer 62.812 MiB
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 1 failed with status 5

Is MTP uses much more memory?

Same issue here: poor token per sec rate & same error messages. M1 Ultra w/ 64GB RAM, so memory is not the primary culprit here.

Tried the Qwen3.6 MTP GGUFs from the MTP branch's author and found they are working fine (although the performance increase seems to be more in the 20% range on Metal compared to 50+% with CUDA GGML backend.)

I guess there is an issue on Metal with Unsloth's GGUFs.

danielhanchen unpinned discussion May 16

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