Instructions to use unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF with Transformers:

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

pipe = pipeline("text-generation", model="unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF")
model = AutoModelForMultimodalLM.from_pretrained("unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF")

llama-cpp-python

How to use unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF",
	filename="DeepSeek-R1-0528-Qwen3-8B-BF16.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:UD-Q4_K_XL
# Run inference directly in the terminal:
llama-cli -hf unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:UD-Q4_K_XL

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:UD-Q4_K_XL
# Run inference directly in the terminal:
llama-cli -hf unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:UD-Q4_K_XL

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/DeepSeek-R1-0528-Qwen3-8B-GGUF:UD-Q4_K_XL
# Run inference directly in the terminal:
./llama-cli -hf unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:UD-Q4_K_XL

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/DeepSeek-R1-0528-Qwen3-8B-GGUF:UD-Q4_K_XL
# Run inference directly in the terminal:
./build/bin/llama-cli -hf unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:UD-Q4_K_XL

Use Docker

docker model run hf.co/unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:UD-Q4_K_XL

LM Studio
Jan

vLLM

How to use unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "unsloth/DeepSeek-R1-0528-Qwen3-8B-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/DeepSeek-R1-0528-Qwen3-8B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:UD-Q4_K_XL

SGLang

How to use unsloth/DeepSeek-R1-0528-Qwen3-8B-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/DeepSeek-R1-0528-Qwen3-8B-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/DeepSeek-R1-0528-Qwen3-8B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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/DeepSeek-R1-0528-Qwen3-8B-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/DeepSeek-R1-0528-Qwen3-8B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF with Ollama:
```
ollama run hf.co/unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:UD-Q4_K_XL
```

Unsloth Studio

How to use unsloth/DeepSeek-R1-0528-Qwen3-8B-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/DeepSeek-R1-0528-Qwen3-8B-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/DeepSeek-R1-0528-Qwen3-8B-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/DeepSeek-R1-0528-Qwen3-8B-GGUF to start chatting

How to use unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:UD-Q4_K_XL

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/DeepSeek-R1-0528-Qwen3-8B-GGUF:UD-Q4_K_XL"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:UD-Q4_K_XL

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/DeepSeek-R1-0528-Qwen3-8B-GGUF:UD-Q4_K_XL

Run Hermes

hermes

Atomic Chat new
Docker Model Runner
How to use unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF with Docker Model Runner:
```
docker model run hf.co/unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:UD-Q4_K_XL
```

Lemonade

How to use unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:UD-Q4_K_XL

Run and chat with the model

lemonade run user.DeepSeek-R1-0528-Qwen3-8B-GGUF-UD-Q4_K_XL

List all available models

lemonade list

Update - Tool Calling + Chat Template bug fixes

pinned

by danielhanchen - opened Jun 9, 2025

Discussion

danielhanchen

Unsloth AI org Jun 9, 2025

•

edited Jun 10, 2025

Just updated DeepSeek-R1-0528-Qwen3-8B GGUFs and BnB, unsloth-BnB quants and all BF16 safetensors.

Native tool calling is now supported. Uses https://github.com/sgl-project/sglang/pull/6765 and https://github.com/vllm-project/vllm/pull/18874 which shows DeepSeek-R1 (not Qwen) getting 93.25% on the BFCL Berkeley Function-Calling Leaderboard https://gorilla.cs.berkeley.edu/leaderboard.html.
Use it via --jinja in llama.cpp. Native transformers and vLLM should work as well.
Had to fix multiple issues in SGLang and vLLM's PRs (dangling newlines etc)
Chat template bug fixes add_generation_prompt now works - previously <|Assistant|> was auto appended - now it's toggle-able. Fixes many issues, and should streamline chat sessions.
UTF-8 encoding of tokenizer_config.json is now fixed - now works in Windows.
Ollama is now fixed on using more memory - I removed num_ctx and num_predict -> it'll now default to Ollama's defaults. This allocated more KV cache VRAM, thus spiking VRAM usage. Please update your context length manually.
[10th June 2025] Update - LM Studio now also works

Please re-download all weights to get the latest updates!

danielhanchen pinned discussion Jun 9, 2025

danielhanchen changed discussion title from June 8th 2025 Update - Tool Calling + Chat Template fixes to June 8th Update - Tool Calling + Chat Template fixes Jun 9, 2025

danielhanchen changed discussion title from June 8th Update - Tool Calling + Chat Template fixes to Update - Tool Calling + Chat Template bug fixes Jun 9, 2025

engrtipusultan

Jun 10, 2025

It is updated again 2 hours ago can you what has changed?

shimmyshimmer

Unsloth AI org Jun 10, 2025

It is updated again 2 hours ago can you what has changed?

Fixed specifically so it has combability for LM Studio because previously it worked in Ollama, llama.cpp etc but not LM Studio

danielhanchen

Unsloth AI org Jun 10, 2025

@engrtipusultan Apologies - no need to re-download if NOT using lm studio - ie llama.cpp, transformers etc are fine.

LM Studio users said our new chat template update didnt work, so I had to redo them.

If you want to be super sure, then you're more than happy to redownload them, but it's not necessary

ideosphere

Jun 10, 2025

Just updated DeepSeek-R1-0528-Qwen3-8B GGUFs and BnB, unsloth-BnB quants and all BF16 safetensors.
...
Please re-download all weights to get the latest updates!

Thanks for the quants & fixes!

BTW I wonder if it'd make sense (in general) to use the '--no-tensor-first-split' option and always make such split GGUFs
so when one changes only the metadata and not the model tensors / weights it's just going to effect a 16kByte first-part file as opposed to a NN GByte first-part file which would have to be re-downloaded in its entirety as you said.

https://github.com/search?q=repo%3Aggml-org%2Fllama.cpp%20--no-tensor-first-split&type=code

"--no-tensor-first-split do not add tensors to the first split (disabled by default)"

danielhanchen

Unsloth AI org Jun 10, 2025

@ideosphere Oh wait someone did mention it to me, (wait found it https://huggingface.co/unsloth/Qwen3-235B-A22B-GGUF/discussions/20) - I accidentally forgot to reply sorry!!!

Yes I like your idea - I'll do it for the larger ones. The small ones like Qwen probs not, since Ollama doesn't like multiple split files

fire

Jun 15, 2025

•

edited Jun 16, 2025

@danielhanchen is there extra configuration needed in ollama to enable tool calling? pretty consistently getting {"error":"hf.co/unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:Q4_K_XL does not support tools"}

( works fine in up to date llama.cpp with jinja, as expected )

edit: hmm, sometimes it crashes llama.cpp with libc++abi: terminating due to uncaught exception of type std::runtime_error: Unexpected empty grammar stack after accepting piece: <｜tool▁calls▁begin｜> ( multiple llama.cpp versions; on a 128gb m4 macbook pro )

mlaihk

Jun 17, 2025

@danielhanchen is there extra configuration needed in ollama to enable tool calling? pretty consistently getting {"error":"hf.co/unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:Q4_K_XL does not support tools"}

Same for me. Just redownloaded Q4_k_XL and ollama still complaining that this model does not support tools.

danielhanchen

Unsloth AI org Jun 18, 2025

@fire For Ollama?
@mlaihk ok that is a weird message for llama.cpp - is this related https://github.com/ggml-org/llama.cpp/issues/13690 ?

fire

Jun 18, 2025

@danielhanchen yes prior to the edit is ollama as noted. after the edit is llama.cpp as noted ( i figured if ollama's template is wrong i would try the other method, but alas )

em-sailfish-cloud

Jun 25, 2025

What's the fix? I'm running into this currently.

davemarr

Jul 14, 2025

Also seeing the error "hf.co/unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:Q4_K_XL does not support tools" when using ollama on osx. Is it wrong for me to assume I could use this model locally without any fancy hardware ( a Mac M1 Pro with built in GPU.. )

leniad

Aug 29, 2025

I still have an issue with this model, in particular in Continue via Ollama, I got an error that this model does not support tools.

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