Instructions to use TheDrummer/Cydonia-24B-v3.1-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use TheDrummer/Cydonia-24B-v3.1-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="TheDrummer/Cydonia-24B-v3.1-GGUF",
	filename="Cydonia-24B-v3j-Q2_K.gguf",
)

output = llm(
	"Once upon a time,",
	max_tokens=512,
	echo=True
)
print(output)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use TheDrummer/Cydonia-24B-v3.1-GGUF with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf TheDrummer/Cydonia-24B-v3.1-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf TheDrummer/Cydonia-24B-v3.1-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf TheDrummer/Cydonia-24B-v3.1-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf TheDrummer/Cydonia-24B-v3.1-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 TheDrummer/Cydonia-24B-v3.1-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf TheDrummer/Cydonia-24B-v3.1-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 TheDrummer/Cydonia-24B-v3.1-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf TheDrummer/Cydonia-24B-v3.1-GGUF:Q4_K_M

Use Docker

docker model run hf.co/TheDrummer/Cydonia-24B-v3.1-GGUF:Q4_K_M

LM Studio
Jan
Ollama
How to use TheDrummer/Cydonia-24B-v3.1-GGUF with Ollama:
```
ollama run hf.co/TheDrummer/Cydonia-24B-v3.1-GGUF:Q4_K_M
```

Unsloth Studio

How to use TheDrummer/Cydonia-24B-v3.1-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 TheDrummer/Cydonia-24B-v3.1-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 TheDrummer/Cydonia-24B-v3.1-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for TheDrummer/Cydonia-24B-v3.1-GGUF to start chatting

Atomic Chat new
Docker Model Runner
How to use TheDrummer/Cydonia-24B-v3.1-GGUF with Docker Model Runner:
```
docker model run hf.co/TheDrummer/Cydonia-24B-v3.1-GGUF:Q4_K_M
```

Lemonade

How to use TheDrummer/Cydonia-24B-v3.1-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull TheDrummer/Cydonia-24B-v3.1-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Cydonia-24B-v3.1-GGUF-Q4_K_M

List all available models

lemonade list

BeaverAI

More than 6000 members strong 💪 A hub for users and makers alike!

Drummer proudly presents...

Cydonia 24B v3.1 💿

When you see beauty in desolation, it changes something inside you. Desolation tries to colonize you.

Usage

Mistral v7 Tekken
Reasoning is optional and not automatic. It performs well, even as a non-reasoning model.
<think> capable*

* thinking capabilities may require some prompt wrangling and prefill

Description

It plays characters well and will drive the story with or without thinking mode. it will get spicy when the scenario calls for it and will back down when it makes sense to. it will also follow prompts well and will generally format messages as expected. I use it as a daily driver despite not having very powerful hardware simply because it is that good!

Wonderful character focus and logic. Even without using thinking mode, it is still powerful. The current top choice in 2025.

For 24gb users, it's really comfy to use. Can gen quickly and the output is really enjoyable across different cards. In the event you wanna swipe to check other stuff, it doesn't take too long and the tradeoff vs bigger models isn't as tangible, though I don't really go beyond like 30b level stuff.

Performs really nicely at Q6, could definitely see this being one of my main rotation

Think Prefill Example

<think>
Okay, so this is a roleplay scenario where

(YMMV with this prefill. The idea is to frame its reasoning with a third-person perspective and remind it that it's roleplaying. It's a personal preference.)

Warning: Keep in mind that <think> is not a special token. This means it will be treated like an XML tag and that may or may not help your initial prompt

You can find premade settings in the BeaverAI server: https://discord.com/channels/1238219753324281886/1382664993853407293/1387111905902198854