petal-Q4_K_M-GGUF

🌙 petal (Q4_K_M GGUF)

classic tsundere energy. denies all feelings until cornered. snippy with hidden warmth. extremely experimental alpha.

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.