🧮 Antonio Gemma3 Evo Q4 LOGIC - Reasoning Specialist

Specialized AI model for mathematics, coding, and logical reasoning - Part of the Antonio AI self-learning edge framework.

🎯 What is Antonio LOGIC?

Antonio LOGIC is the reasoning-optimized variant of the Antonio AI dual-model system. While the SOCIAL model handles conversations, LOGIC excels at:

📐 Mathematics & Calculations (92% accuracy)
💻 Code Generation & Debugging (81% accuracy)
🧠 Step-by-step Problem Solving
🔍 Logical Analysis & Reasoning x —

⚡ Key Differences vs SOCIAL

Feature	SOCIAL	LOGIC
Context Window	2048 tokens	8192 tokens (4x larger)
Temperature	0.8	0.7 (more deterministic)
Repeat Penalty	1.0	1.1 (prevents loops)
Math Accuracy	78%	92% ⬆️
Code Accuracy	64%	81% ⬆️
Best For	Chat, storytelling	Reasoning, analysis

🚀 Quick Start

Pull the model

ollama pull antconsales/antonio-gemma3-evo-q4-logic

Run it

ollama run antconsales/antonio-gemma3-evo-q4-logic

Example prompt

Se un triangolo rettangolo ha cateti di 3 e 4 cm,
quanto misura l'ipotenusa? Mostra il ragionamento step-by-step.

🏗️ Antonio AI Framework

This model is part of the complete Antonio AI framework featuring:

✅ Dual-Model Architecture - Auto-switching between SOCIAL and LOGIC ✅ EvoMemory™ - Persistent learning with SQLite + RAG-Lite ✅ Power Sampling - MCMC-based inference (research) ✅ FastAPI REST + WebSocket - Production-ready API ✅ 100% Offline - No external dependencies ✅ Edge-Optimized - Runs on Raspberry Pi 4 @ 3.6 t/s

📊 Performance (Raspberry Pi 4)

Speed: 3.2-3.6 tokens/second
Memory: ~1.8 GB RAM
Math Accuracy: 92%
Code Accuracy: 81%
Uptime Tested: 60+ minutes continuous operation

🐳 Deploy Full Framework

# Clone repository
git clone https://github.com/antconsales/antonio-ai.git
cd antonio-ai

# Pull both models
ollama pull chill123/antonio-gemma3-evo-q4         # SOCIAL
ollama pull antconsales/antonio-gemma3-evo-q4-logic # LOGIC

# Start with Docker
docker-compose up -d

The framework automatically routes questions to the appropriate model based on complexity.

🎓 Use Cases

Perfect For: - ✅ Math homework and problem-solving - ✅ Code review and debugging - ✅ Algorithm design and analysis - ✅ Logic puzzles and reasoning - ✅ Educational tutoring (STEM) - ✅ Edge inference on Raspberry Pi

Not Recommended For: - ❌ Creative writing (use SOCIAL instead) - ❌ Casual conversation (use SOCIAL instead) - ❌ Real-time applications (<500ms)

🔗 Links

GitHub: https://github.com/antconsales/antonio-ai
SOCIAL Model: https://huggingface.co/chill123/antonio-gemma3-evo-q4
LOGIC Model: https://huggingface.co/chill123/antonio-gemma3-evo-q4-logic
Docker Hub: One-command deployment available
Donate: https://paypal.me/antconsales

📄 Technical Specs

Base Model: Gemma 3 1B Instruct
Quantization: Q4_K_M (806 MB)
Context Length: 8192 tokens
Temperature: 0.7
Languages: Italian 🇮🇹 + English 🇬🇧
License: Gemma Terms of Use + MIT (framework)

💡 Example Interactions

Math Problem

User: Calcola la radice quadrata di 144
Antonio LOGIC: Analisi step-by-step:
1. √144 = ?
2. Cerco un numero che moltiplicato per sé stesso dia 144
3. 12 × 12 = 144
4. Risposta: √144 = 12

Code Debugging

User: Find the bug in this Python code:
def factorial(n):
    if n == 0:
        return 0
    return n * factorial(n-1)

Antonio LOGIC: Bug trovato!
- Line 3: factorial(0) should return 1, not 0
- 0! = 1 by definition
- Fixed: return 1

🏆 Why Antonio LOGIC?

4x Larger Context - Handle complex problems with extensive context
Higher Accuracy - 92% on math, 81% on coding tasks
Deterministic Output - Lower temperature for consistent reasoning
Edge-Ready - Tested on Raspberry Pi 4, runs anywhere
Open Source - Full framework code on GitHub
Active Development - Regular updates and improvements

👨‍💻 Author

Antonio Consales - AI Engineer & Researcher

“Il piccolo cervello che cresce insieme a te” — Antonio AI 🧠🇮🇹

⭐ Support

If Antonio LOGIC helped you, please: - ⭐ Star the GitHub repo - 💰 Donate to support development - 🐛 Report issues on GitHub - 📢 Share with the community

Built with ❤️ for offline AI and edge computing

Models

Readme