Phil2Sat/Goetia-26B-A4B-v1.3-Absolute-Heretic-Restored-I-Compact-MXFP4.gguf overview
Restored global attention from: https://huggingface.co/MuXodious/gemma 4 26B A4B it SOMPOA heresy into https://huggingface.co/Naphula/Goetia 26B A4B v1.3 Absol…
Runs locally from ~227.8 MB disk (4 GB VRAM class GPUs with llama.cpp / guIDE).
Repository Files & Downloads
Model Details
Model README
---
base_model:
- Naphula/Goetia-26B-A4B-v1.3-Absolute-Heretic-ARA
---
Restored global attention from:
https://huggingface.co/MuXodious/gemma-4-26B-A4B-it-SOMPOA-heresy
into
https://huggingface.co/Naphula/Goetia-26B-A4B-v1.3-Absolute-Heretic-ARA
It seems it fixed the no-thinking issue
https://huggingface.co/Naphula/Goetia-26B-A4B-v1.3-Absolute-Heretic-ARA/discussions/4
Also it's a custom Frankenquant, a modified version of APEX-Script to get it as small as possible and maintaining context, adherance, intelligence.
At least all my initial tests give perfect german output without any misspelling or wrong written words/Kanji/own bugs or repetition in thinking:
Sampler settings:
- temp: 0.8
- top-p: 0.95
- top-k 64
- nothing else all default / zero
Extra request parameters:
{"chat_template_kwargs":{"enable_thinking": true,"reasoning_effort": "medium","preserve_thinking": false}}Run Phil2Sat/Goetia-26B-A4B-v1.3-Absolute-Heretic-Restored-I-Compact-MXFP4.gguf with guIDE
Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.
Source: Hugging Face · Compare models