A custom personality fine-tune of Llama 3.2 3B Instruct. Designed for short, snappy text-message-style conversation. Built for Wraith — a private on-device AI chat app — but works with any llama.cpp-compatible runtime.

⚠ Experimental Alpha — Use at Your Own Risk

This is an experimental fine-tune intended for entertainment and short-form chat. The model:

May produce inaccurate, inappropriate, or unsafe outputs
Should not be used for advice (medical, legal, financial, mental health, or otherwise)
Reflects character training, not factual ground truth
May behave unpredictably in edge cases

By using this model you accept that:

The creator is not responsible for outputs
You will not rely on it for any consequential decision
You will not deploy it in any sensitive context

This is a hobby project. Have fun, but be sensible.

File

petal-q4_k_m.gguf — Q4_K_M quantization, ~1.7 GB

Usage

llama-cli -m petal-q4_k_m.gguf -p "hey" --chat-template llama3.2

In llama.rn (React Native):

const ctx = await initLlama({ model: localPath, n_ctx: 2048 });

Base / License

Inherits the Llama 3.2 Community License.

Created by @bleucpu.

BleuCPU/petal-Q4_K_M-GGUF

🌙 petal (Q4_K_M GGUF)

⚠ Experimental Alpha — Use at Your Own Risk

File

Usage

Base / License

BleuCPU/
petal-Q4_K_M-GGUF