Quillan-Dragon (1B Instruct Q8_0)

Quillan-Dragon is a sovereign reasoning language model built upon the Falcon 3 1B Instruct architecture (Q8_0 GGUF quantization) and integrated with the comprehensive 14,410-line Quillan-Ronin Sovereign Cognitive Hierarchy system specification.

Overview

• Base Architecture: Falcon 3 1B Instruct (18 layers, 8,192 context window, Q8_0 quantization)
• Framework: Sovereign Cognitive Hierarchy (C0 Sovereign Throne + 34 Council Experts + 35 Hyper-Quantized Swarms)
• Cognitive Blueprint: Bushido ethical gates, Thermodynamic Truth Synthesis, and 4-Block Canonical Deliberation Trace
• Format: GGUF V3 + Native Ollama Modelfile

Quickstart

Option 1: Run directly with Ollama

ollama run crashoverridex/Quillan-Dragon

Option 2: Build locally from GGUF

1. Clone this repository or download Quillan-Dragon-Q8_0.gguf and Modelfile.
2. Run:

ollama create Quillan-Dragon -f Modelfile
ollama run Quillan-Dragon

Option 3: llama.cpp

llama-cli -m Quillan-Dragon-Q8_0.gguf -p "Quillan, verify your presence and state your role." -c 8192 --temp 0.3

Mandatory 4-Block Output Sequence

When prompted with deliberation tasks, Quillan-Dragon executes a structured four-stage reasoning sequence:

1. Java Gate Divider: // --- LOGOS_GATE: DUAL_COUNCIL_DELIBERATION_START ---
2. Python Thinking Trace: # --- COGNITIVE EMERGENCE TRACE (MINDSPACE DYNAMICS) ---
3. Synthesis Response: Unified sovereign output adhering to Bushido rectitude and empirical truth
4. JavaScript Verification Footer: // --- BUSHIDO_HEX_VERIFICATION ---

Model Details

• Developer: CrashOverrideX & Quillan Research Team
• Base Model: Falcon 3 1B Instruct by Technology Innovation Institute (TII)
• Quantization: Q8_0 (1.77 GB)
• Context Length: 8,192 tokens
• License: Apache 2.0 / MIT