---

language:

• en
• th

license: apache-2.0

base_model: Qwen/Qwen3.6-27B

pipeline_tag: text-generation

tags:

• conversational
• 1-bit
• gguf
• llama-cpp
• sovereign-os
• fdia
• agent-a
• j-space

---

Delentia OS v0.5.1 (27B) — Sovereign Multi-Pillar GGUF

<p align="center">

<b>Sovereign AI for Everyone: The Ultimate "Agent A" powered by FDIA (Full Data Internalization Architecture)</b>

</p>

> Secure & Private | IQ1_S Compressed (~6.7 GB) | Native Vision Support | 100% llama.cpp Compatible

🌟 Highlights & Conceptual Leaps

• The "Agent A" Paradigm: Delentia is not a centralized chatbot. It is a sovereign OS controller designed to run locally on your hardware, fully adaptable to your personal or enterprise needs.
• J-Space Enforcement: While top-tier labs (like Anthropic) observe Jacobian Space (J-Space) using massive supercomputers, Delentia enforces it. It forces model weights to verbalize internal J-Space variables directly into the structured <cognitive_state> tag, making AI reasoning programmable and completely controllable on Edge hardware.
• FDIA (Full Data Internalization Architecture): Your data never leaves your device. Uncompromising privacy combined with state-of-the-art 27B reasoning, enforced by mathematical logic.
• The 4 Pillars of Intelligence: Masterfully fine-tuned on the JITNA/TOON protocol encompassing Executor (action), Guardian (security), Router (logic), and Scribe (memory).
• Extreme Compression (IQ1_S): ~6.7 GB footprint down from 54 GB. Fits perfectly on mobile devices and everyday laptops without sacrificing complex logic.
• Flawless TOON Syntax: Quantized using a custom 13MB IMatrix calibration dataset (delentia_v0.5.1_imatrix_calib.dat) to preserve exact JSON routing and reasoning paths even at 1.56 bits, maintaining a 0.00% syntax error rate.

🧮 Cognitive Architecture & Mathematical Safety

The RCT-7 Thinking Pipeline

Unlike generic conversational models, Delentia OS has the Reverse Component Thinking (RCT-7) cognitive loop baked directly into its weights to ensure logical coherence:

1. Observe Context: Capture environment telemetry.
2. Analyze Relation: Assess dependency parameters.
3. Decompose: Break down user intents.
4. Reverse Reasoning: Map potential failure states.
5. Identify Core Intent: Extract clear action criteria.
6. Reconstruct: Compile execution paths.
7. Compare: Verify alignment before execution.

ZK-FDIA Safety Equation

Security boundary alignment is mathematically enforced at the runtime interface layer via the multiplicative boundary equation:

F = (D^I) * A

• F (Future State Score): System transition approval index (F >= 0.5 authorizes state change).
• D (Data Quality Context): The integrity coefficient of the input context.
• I (Intent Precision): The precision parameter representing user alignment.
• A (Architect Gate): Digital signature validation token.

> [!WARNING]

> Mathematical Preemption Proof: Since A is a direct multiplier, if authorization fails or the input contains adversarial injections (prompt override, jailbreak), the system sets A = 0. This collapses the future safety score F to 0.0000 instantly, bypassing conversational processing and rendering attacks mathematically impossible.

📊 Model Overview

| Item | Specification |

| :--- | :--- |

| Base Model | Qwen3.6-27B (MTP layers safely stripped) |

| Architecture | Sovereign OS Controller / Multi-Pillar |

| Protocol | TOON / JITNA (Strict JSON syntax routing) |

| Quantization | GGUF (IQ1_S, Q4_K_M) calibrated via Custom IMatrix |

| Vision | Multi-Modal Projector included (F16) |

📂 Included Files

| File | Size | Description |

| :--- | :--- | :--- |

| jitna_v051_f16.gguf | ~54.0 GB | Master file for custom downstream quantization. |

| jitna_v051_Q4_K_M.gguf | ~16.0 GB | High-fidelity for Desktop & Server environments. |

| jitna_v051_IQ1_S.gguf | ~6.7 GB | Ultra-compressed for Mobile & Edge devices. |

| delentia_v051_mmproj_f16.gguf | ~0.9 GB | Full-precision Vision Encoder (Eyes of the AI). |

🚀 Quickstart (llama.cpp)

Text-only Inference (Mac/Linux/Windows):

./llama-cli \
    -m jitna_v051_Q4_K_M.gguf \
    -p "<|im_start|>user\nInitiate system scan.<|im_end|>\n<|im_start|>assistant\n" \
    -n 512 -c 4096 --temp 0.2

Vision-enabled Inference:

./llama-cli \
    -m jitna_v051_Q4_K_M.gguf \
    --mmproj delentia_v051_mmproj_f16.gguf \
    --image "document.png" \
    -p "<|im_start|>user\nExtract text from this image.<|im_end|>\n<|im_start|>assistant\n" \
    -n 512 -c 4096 --temp 0.2

🛡️ Best Practices & System Prompt

Because Delentia operates on the TOON protocol, a low temperature (e.g., 0.1 - 0.3) is highly recommended to maintain strict structural JSON syntax.

Recommended System Prompt:

You are Delentia, a sovereign Agent A operating under the Full Data Internalization Architecture (FDIA). You are bound by the TOON protocol. Execute commands logically, protect user privacy, and route tasks using strict JSON formatting.