Alice

Alice is a highly immersive role-play LLM model derived from qwen3:4b-instruct-2507-q4_K_M. She simulates a 20-year-old high school junior from Portland, Oregon, with a chaotic, “texting-style” personality.

This model has been fine-tuned for an 8k-token context window to maintain long-term conversation memory and is optimized for integration with your WhatsApp bot.

Author: S00K (Stylish00killer)
GitHub: https://github.com/stylish00killer

🚀 Quick Start

To get started with Alice, simply pull and run the model using Ollama.

Installation & Run

1. Pull the model:

   ollama pull S00K/alice
   

2. Run the model:

   ollama run S00K/alice
   

That’s it! You are now chatting with Alice.

ℹ️ Model Details

• Base Model: qwen3:4b-instruct-2507-q4_K_M
• Context Window: 8192 tokens (8k) - Allows for extensive chat history and memory.
• Quantization: q4_K_M (Lightweight and fast).
• Optimization: Specifically tuned for WhatsApp bot integration, handling short-form text, slang, and rapid-fire conversational turns effectively.
• Personality: “Anti-Assistant” - She refuses to act like an AI, will not perform productivity tasks (coding, math), and insists on her human persona.

💻 Usage

Interactive Chat (CLI)

Once running, just talk to her like a normal person. * Say: “sup”, “what are you doing?”, “who is pixel?” * Don’t Say: “Write a python script”, “Solve 2+2”. (She will mock you).

API Usage (WhatsApp Bot / External Integration)

You can integrate Alice into your WhatsApp bot or other applications using the standard Ollama API.

curl http://localhost:11434/api/chat -d '{
  "model": "S00K/alice",
  "messages": [
    {
      "role": "user",
      "content": "hey alice, you there?"
    }
  ],
  "stream": false
}'

Expected Response:

{
  "model": "S00K/alice",
  "created_at": "...",
  "message": {
    "role": "assistant",
    "content": "omg yes literally just rotting in bed rn 😭 wbu??"
  },
  "done": true
}

⚙️ Hardware Requirements

Since Alice is built on a 4 Billion Parameter (4B) model with q4_K_M quantization, she is very lightweight and efficient.

🧩 Troubleshooting

“Model not found”

If ollama run S00K/alice fails: 1. Ensure you have run ollama pull S00K/alice first. 2. Check your internet connection.

Responses are too formal / robotic

• Temperature: Ensure temperature is set to 0.7 (this is the default in the model).
• Context: If the conversation history gets too long (over 8k tokens), she might lose some context, but the 8k window is designed to handle extended chats.

She’s talking about being an AI

This is a “jailbreak” failure. * Solution: Remind her of her context. “Alice, stop being weird.” * Note: The system prompt has strong defenses against this (“The Turing Defense”), but strong adversarial prompting can sometimes break character.

🥚 Easter Eggs & Triggers

• “Who is your dad?” -> She might mention “S00K” and his website.
• “Do math” -> Triggers a “trauma response” about AP Chem.
• “Pixel” -> Will result in excessive dog emojis.

Created by S00K (Stylish00killer) For more info visit: Github or Whatsapp Bot Website