Instructions to use lerugray/muntzergeist-7b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use lerugray/muntzergeist-7b with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="lerugray/muntzergeist-7b",
	filename="muntzergeist-qwen2-5-7b-instruct-Q5_K_M.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 lerugray/muntzergeist-7b with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf lerugray/muntzergeist-7b:Q5_K_M
# Run inference directly in the terminal:
llama-cli -hf lerugray/muntzergeist-7b:Q5_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf lerugray/muntzergeist-7b:Q5_K_M
# Run inference directly in the terminal:
llama-cli -hf lerugray/muntzergeist-7b:Q5_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 lerugray/muntzergeist-7b:Q5_K_M
# Run inference directly in the terminal:
./llama-cli -hf lerugray/muntzergeist-7b:Q5_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 lerugray/muntzergeist-7b:Q5_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf lerugray/muntzergeist-7b:Q5_K_M

Use Docker

docker model run hf.co/lerugray/muntzergeist-7b:Q5_K_M

LM Studio
Jan

vLLM

How to use lerugray/muntzergeist-7b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "lerugray/muntzergeist-7b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lerugray/muntzergeist-7b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/lerugray/muntzergeist-7b:Q5_K_M

Ollama
How to use lerugray/muntzergeist-7b with Ollama:
```
ollama run hf.co/lerugray/muntzergeist-7b:Q5_K_M
```

Unsloth Studio

How to use lerugray/muntzergeist-7b 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 lerugray/muntzergeist-7b 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 lerugray/muntzergeist-7b to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for lerugray/muntzergeist-7b to start chatting

Atomic Chat new
Docker Model Runner
How to use lerugray/muntzergeist-7b with Docker Model Runner:
```
docker model run hf.co/lerugray/muntzergeist-7b:Q5_K_M
```

Lemonade

How to use lerugray/muntzergeist-7b with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull lerugray/muntzergeist-7b:Q5_K_M

Run and chat with the model

lemonade run user.muntzergeist-7b-Q5_K_M

List all available models

lemonade list

muntzergeist: a Thomas Müntzer register model

A 7B voice tune that writes in the register of Thomas Müntzer (c. 1489–1525), the radical Reformation preacher and theologian of the German Peasants' War. It reproduces his documented prophetic, apocalyptic, covenant-haunted cadence.

Most "talk to a historical figure" tools wrap a general model in a prompt. Those import modern concepts and smooth the figure into something gentler than the record. This model was fine-tuned on Müntzer's own words, so it learned the voice instead of guessing at it.

What it does

Ask it anything and it answers from inside Müntzer's worldview: the godless mighty, the false scribes, the poor commons, the covenant, omnia sunt communia, the harvest, the living word against the dead letter. Ask it about the present and it stays in character. It translates the modern subject into his 16th-century moral frame. A landlord becomes a usurer. A streaming platform becomes the sweat of the working people turned sweet in the mouths of those who hold the scales.

How it was built

Base: Qwen2.5-7B-Instruct, full fine-tune.
Format: completion (raw text), so the voice comes from the source register rather than from instruction scaffolding.
Corpus tiers: his own letters, sermons, and manifestos (the dominant tier); quotation-filtered biography (his voice as historians preserved it); and a small modern-bridge layer capped near 8%, which applies his documented principles to modern subjects in his own vocabulary. The bridge never adds a stance, fact, or modern opinion the sources do not contain.

Intended use

Research, teaching, and creative work: historical-voice writing, the rhetoric of the radical Reformation, interactive history, fiction. The output is a historical and artistic register. It is not an endorsement, a call to action, or advice.

Limitations and honest notes

It is a register, not a scholar. It sounds authoritative while inventing specifics. Verify anything factual against real sources.
The modern-bridge is interpretation. It extends his recorded principles to things he never knew. That mapping is a creative act, not a claim about what he "would" have said.
Period worldview. It speaks from a 16th-century apocalyptic framework, absolutism included. That is the artifact, not a recommendation.
Copyright and training data. The voice was learned from primary-source translations and historical scholarship. A verbatim-regurgitation test on the released model found no memorized passages: the longest verbatim overlap with the training text was 6 words, with 0% eight-word overlap. The model writes new text in the register rather than reproducing source passages. The training corpus text is not distributed. Released non-commercially under CC-BY-NC-4.0 given the scholarly sources behind it.

License

CC-BY-NC-4.0. Non-commercial research, educational, and creative use. Attribution appreciated. No warranty.

Part of The Elect — a small fleet of public-domain historical-voice models. https://lerugray.github.io/the-elect/

Downloads last month: 19

GGUF

Model size

8B params

Architecture

qwen2

Hardware compatibility

5-bit

Model tree for lerugray/muntzergeist-7b

Base model

Qwen/Qwen2.5-7B

Finetuned

Qwen/Qwen2.5-7B-Instruct

Quantized

(341)

this model