paperboygold/gpt-oss-sanguine-20b-q4-gguf

Sanguine Scribe Q4 GGUF

This is a Q4 quantized version of gpt-oss-sanguine-20b-v1, a consequence-based alignment model for character roleplay.

• File size: 15.8 GB
• Quantization: Q4
• Format: GGUF (llama.cpp compatible)

Usage with llama.cpp

# Download the model
huggingface-cli download paperboygold/sanguine-scribe-q4-gguf sanguine_scribe_q4_k_m_20250818_140934.gguf --local-dir ./

# Run with llama.cpp
./llama-cli -m sanguine_scribe_q4_k_m_20250818_140934.gguf -p "You are a tavern keeper. A hooded stranger approaches."

Original Model Details

Sanguine Scribe implements consequence-based alignment for character roleplay:

• Base Model: openai/gpt-oss-20b
• Training Dataset: 350,969 examples from myriad sources
• Training Loss: 4.1 → 1.31 over 500 steps
• Approach: Realistic consequences instead of refusal responses

Perfect for immersive character roleplay and interactive storytelling.