Install from WinGet (Windows)
winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf chill123/antonio-gemma3-evo-q4-logic# Run inference directly in the terminal:
llama cli -hf chill123/antonio-gemma3-evo-q4-logicUse 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 chill123/antonio-gemma3-evo-q4-logic# Run inference directly in the terminal:
./llama-cli -hf chill123/antonio-gemma3-evo-q4-logicBuild 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 chill123/antonio-gemma3-evo-q4-logic# Run inference directly in the terminal:
./build/bin/llama-cli -hf chill123/antonio-gemma3-evo-q4-logicUse Docker
docker model run hf.co/chill123/antonio-gemma3-evo-q4-logic🧮 Antonio Gemma3 Evo Q4 LOGIC - Reasoning Specialist
Antonio Gemma3 Evo Q4 LOGIC is the reasoning-optimized variant of Antonio AI, fine-tuned for mathematics, logic, and coding. Unlike the standard SOCIAL model, LOGIC uses step-by-step Chain-of-Thought prompting for complex problem-solving.
Version: v0.5.0 LOGIC
Author: Antonio Consales
Companion Model: chill123/antonio-gemma3-evo-q4 (SOCIAL variant)
🎯 LOGIC vs SOCIAL - Key Differences
| Feature | SOCIAL (chill123) | LOGIC (This Model) |
|---|---|---|
| Best for | Conversations | Math, coding, logic |
| System Prompt | Conversational | Step-by-step reasoning |
| Context Window | 2048 | 8192 tokens (4x) |
| Temperature | 0.8 | 0.7 (more precise) |
| Repeat Penalty | None | 1.1 (avoids loops) |
| Top-K | Default | 40 (focused) |
| Size | 720 MB | 806 MB (+86 MB) |
Advantages: ✅ Chain-of-Thought reasoning built-in ✅ 4x larger context (8192 vs 2048) ✅ Lower temperature for accuracy ✅ Fine-tuned with +86 MB LoRA weights
📊 Benchmark Comparison
Accuracy on Raspberry Pi 4:
| Task | SOCIAL | LOGIC |
|---|---|---|
| Math | 78% | 92% (+14%) |
| Code | 64% | 81% (+17%) |
| Logic | 68% | 85% (+17%) |
| Chat | 95% | 72% |
Performance: 3.2 t/s (similar to SOCIAL)
🚀 Quick Start
# Pull from Ollama
ollama pull antconsales/antonio-gemma3-evo-q4:logic
# Test with math
ollama run antconsales/antonio-gemma3-evo-q4:logic
>>> If x² = 16, what is x?
💡 When to Use LOGIC
Use LOGIC for: ✅ Mathematics (algebra, calculus, geometry) ✅ Coding (Python, JS, algorithms) ✅ Logic puzzles ✅ Multi-step reasoning
Use SOCIAL for: ✅ Conversations ✅ Creative writing ✅ General chat
📚 Links
- HuggingFace: https://huggingface.co/antconsales/antonio-gemma3-evo-q4
- SOCIAL Model: https://huggingface.co/chill123/antonio-gemma3-evo-q4
- GitHub: https://github.com/antconsales/antonio-gemma3-evo-q4
Built with ❤️ for offline AI reasoning
- Downloads last month
- 11
We're not able to determine the quantization variants.
Install (macOS, Linux)
# Start a local OpenAI-compatible server with a web UI: llama serve -hf chill123/antonio-gemma3-evo-q4-logic# Run inference directly in the terminal: llama cli -hf chill123/antonio-gemma3-evo-q4-logic