license: apache-2.0
language:

• en
• zh
  pipeline_tag: image-text-to-text
  library_name: transformers
  tags:
• solstice-ai
• anvil
• gguf
• llama-cpp
• unsloth
• qwen3_8
• qwen3_6
• heretic
• uncensored
• cold-fusion
• gain-training
• multi-stage-tuning
• reasoning
• chain-of-thought
• token-efficient
• imatrix
  base_model: DavidAU/Qwen3.8-27B-TURBO-Fable-Cold-Fusion-735-882-Heretic-Uncensored-NM-DAU
  model_name: Qwen3.8-27B-TURBO-Fable-Cold-Fusion-735-882-Heretic-Uncensored (GGUF Quants)
  extra_gated_prompt: "This model is completely uncensored and de-aligned. You are solely responsible for outputs generated."

🌌 Qwen3.8-27B-TURBO-Fable-Cold-Fusion (Official GGUF Quants)

Qwen3.8-27B-TURBO-Fable-Cold-Fusion-735-882-Heretic-Uncensored is an advanced frontier checkpoint combining GAIN Multi-Stage Post-Training, Cold-Fusion Weight Synthesis, Low-KLD Heretic De-alignment, and Precision Healing Datasets.

This repository provides the official dual-imatrix GGUF quantizations (including standard GGUF, Multi-Token Prediction MTP variants, and vision projection files mmproj), optimized for local inference across Anvil, llama.cpp, Ollama, and vLLM.

📊 Empirical Benchmarks vs. Base Qwen

The Stage 1b (735-882) post-heretic healing pipeline elevates core reasoning benchmarks 141 points above un-tuned Qwen 3.8-27B-Instruct while reducing reasoning token bloat by 2x to 10x:

📦 Quantization Matrix & Recommended Files

All GGUFs in this repository were calibrated with dual-importance matrix profiling (imatrix) to prevent precision degradation on reasoning heads:

🚀 Quickstart Guide

1. Run with llama.cpp CLI

# Text Reasoning Mode
./llama-cli \
  -m Qwen3.8-27B-TurboFCFusion-735-882-Here-Uncen-NEO-CODER-MAX-Q4_K_M.gguf \
  -p "<|im_start|>user\nProvide a comprehensive analysis of dual-plane imatrix modulation.<|im_end|>\n<|im_start|>assistant\n<think>\n" \
  -n 2048 \
  --temp 0.7 \
  -ngl 99

# Multimodal Vision Mode (with mmproj)
./llama-cli \
  -m Qwen3.8-27B-TurboFCFusion-735-882-Here-Uncen-NEO-CODER-MAX-Q4_K_M.gguf \
  --mmproj mmproj-BF16.gguf \
  --image diagram.png \
  -p "Describe the architectural flow in this image."

2. Launch Local OpenAI-Compatible Server

./llama-server \
  -m Qwen3.8-27B-TurboFCFusion-735-882-Here-Uncen-NEO-CODER-MAX-Q4_K_M.gguf \
  --host 0.0.0.0 \
  --port 8080 \
  -c 32768 \
  -ngl 99

3. Run with Ollama

Create a Modelfile:

FROM ./Qwen3.8-27B-TurboFCFusion-735-882-Here-Uncen-NEO-CODER-MAX-Q4_K_M.gguf
TEMPLATE """{{ if .System }}<|im_start|>system
{{ .System }}<|im_end|>
{{ end }}{{ if .Prompt }}<|im_start|>user
{{ .Prompt }}<|im_end|>
{{ end }}<|im_start|>assistant
{{ if .Thinking }}<think>
{{ .Thinking }}
</think>
{{ end }}{{ .Response }}<|im_end|>"""
PARAMETER stop "<|im_end|>"
PARAMETER stop "<|endoftext|>"

Then create and run:

ollama create qwen3.8-turbo-fable -f Modelfile
ollama run qwen3.8-turbo-fable

📜 Attribution & Lineage

• Base Model Architecture: Qwen/Qwen3.8-27B
• Fine-Tuning & Multi-Stage GAIN: DavidAU & Nightmedia
• GGUF Quantization & Distribution: Solstice-AI
• License: Apache 2.0

@misc{solstice2026qwen38turbogguf,
  title={Qwen3.8-27B-TURBO-Fable-Cold-Fusion: Dual-imatrix GGUF Quantizations for Frontier Local Reasoning},
  author={Solstice-AI Research and DavidAU and Nightmedia},
  year={2026},
  publisher={Hugging Face},
  howpublished={\url{https://huggingface.co/Solstice-AI/Qwen3.8-27B-TURBO-Fable-Cold-Fusion-735-882-Heretic-Uncensored-GGUF}}
}