Instructions to use unsloth/Devstral-2-123B-Instruct-2512-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- llama-cpp-python
How to use unsloth/Devstral-2-123B-Instruct-2512-GGUF with llama-cpp-python:
# !pip install llama-cpp-python from llama_cpp import Llama llm = Llama.from_pretrained( repo_id="unsloth/Devstral-2-123B-Instruct-2512-GGUF", filename="BF16/Devstral-2-123B-Instruct-2512-BF16-00001-of-00006.gguf", )
llm.create_chat_completion( messages = "No input example has been defined for this model task." )
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use unsloth/Devstral-2-123B-Instruct-2512-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/Devstral-2-123B-Instruct-2512-GGUF:UD-Q4_K_XL # Run inference directly in the terminal: llama cli -hf unsloth/Devstral-2-123B-Instruct-2512-GGUF:UD-Q4_K_XL
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf unsloth/Devstral-2-123B-Instruct-2512-GGUF:UD-Q4_K_XL # Run inference directly in the terminal: llama cli -hf unsloth/Devstral-2-123B-Instruct-2512-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/Devstral-2-123B-Instruct-2512-GGUF:UD-Q4_K_XL # Run inference directly in the terminal: ./llama-cli -hf unsloth/Devstral-2-123B-Instruct-2512-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/Devstral-2-123B-Instruct-2512-GGUF:UD-Q4_K_XL # Run inference directly in the terminal: ./build/bin/llama-cli -hf unsloth/Devstral-2-123B-Instruct-2512-GGUF:UD-Q4_K_XL
Use Docker
docker model run hf.co/unsloth/Devstral-2-123B-Instruct-2512-GGUF:UD-Q4_K_XL
- LM Studio
- Jan
- Ollama
How to use unsloth/Devstral-2-123B-Instruct-2512-GGUF with Ollama:
ollama run hf.co/unsloth/Devstral-2-123B-Instruct-2512-GGUF:UD-Q4_K_XL
- Unsloth Studio
How to use unsloth/Devstral-2-123B-Instruct-2512-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/Devstral-2-123B-Instruct-2512-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/Devstral-2-123B-Instruct-2512-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/Devstral-2-123B-Instruct-2512-GGUF to start chatting
- Pi
How to use unsloth/Devstral-2-123B-Instruct-2512-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/Devstral-2-123B-Instruct-2512-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/Devstral-2-123B-Instruct-2512-GGUF:UD-Q4_K_XL" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Hermes Agent new
How to use unsloth/Devstral-2-123B-Instruct-2512-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/Devstral-2-123B-Instruct-2512-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/Devstral-2-123B-Instruct-2512-GGUF:UD-Q4_K_XL
Run Hermes
hermes
- Atomic Chat new
- OpenClaw new
How to use unsloth/Devstral-2-123B-Instruct-2512-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/Devstral-2-123B-Instruct-2512-GGUF:UD-Q4_K_XL
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/Devstral-2-123B-Instruct-2512-GGUF:UD-Q4_K_XL" \ --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/Devstral-2-123B-Instruct-2512-GGUF with Docker Model Runner:
docker model run hf.co/unsloth/Devstral-2-123B-Instruct-2512-GGUF:UD-Q4_K_XL
- Lemonade
How to use unsloth/Devstral-2-123B-Instruct-2512-GGUF with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull unsloth/Devstral-2-123B-Instruct-2512-GGUF:UD-Q4_K_XL
Run and chat with the model
lemonade run user.Devstral-2-123B-Instruct-2512-GGUF-UD-Q4_K_XL
List all available models
lemonade list
mmproj?
am i just blind? i don't see the file??
Currently llama.cpp only supports text only. Vision is been worked on I think
Appreciate the response @danielhanchen but it's a bit confusing because I've been running --mmproj with Qwen3-VL-30B-A3B-Instruct-UD-Q8_K_XL. Using the param --mmproj /mnt/data/models/Qwen3-VL-30B-A3B-Instruct-UD-Q8_K_XL/mmproj-BF16.gguf the model does load correctly and even in the stripped down llama.cpp web ui, I can paste in images to the chat with no issues.
@aaron-newsome are you in the right repo? Or you mean you used the Qwen3-VL mmproj for Devstral-2-123B?
my question was, where is the mmproj file for Devstral-2-123B-Instruct-2512-GGUF
then danielhanchen said llama.cpp doesn't support images
to which i responded, sure it does because i've used images with llama.cpp and other image models
i'm in the right place
Oh I see! No he was meaning not supported yet for this specific model. Sorry for the misunderstanding! I thought you tried to load the Qwen-VL mmproj alongside Devstral-2-123b π
If someone can provide instructions for extracting the encoder and converting it to .mproj, Iβm willing to execute them. Let me know the steps.
Maybe we should ask the guys who made the Qwen3-VL-30B-A3B-Instruct-UD-Q8_K_XL/mmproj-BF16.gguf file and see how they made it? I'd be happy to give it a try myself but I have no idea where to start. I also think that @danielhanchen is suggesting that it's not really about making the mmproj file itself, but that llama.cpp needs to support it. I can definitely say the Qwen3-VL works but I have no idea whether llama.cpp would support the mmproj file for this model if it existed. Looking at the llama.cpp guides, it doesn't really say that only mmproj files are supported for certain models, it seems generic like any model that has an mmproj file would be supported. Again, I have no idea.
@dr-e deepwiki is really cool you should try it, it helps really much when you need to dig into documentation ;)
https://deepwiki.com/search/how-to-create-the-mmprojgguf-f_461ddcaa-3ec8-41c3-b19c-cbac84264874
According to unsloth own website the mmproj file should exist in this repository, but it does not exists! β¦ https://unsloth.ai/docs/models/devstral-2
In my opinion this a mistake on unsloth side of things. If its a documentation issue or an issue with the upload to the repo is on unsloth to figure out.
Yep, I think the unsloth docs are incorrect, copied from another model (like devstral small 2).
It seems that only devstral small 2 supports vision. That SORT OF makes sense as it's smaller and faster for ongoing work like checking lots of screen updates or processing lots of images, and you'd normally use a larger model for more "thoughtful" stuff.