Gemma 3 4b It MAX HORROR Imatrix GGUF
A horror-themed specialized large language model enhanced with Neo Horror Imatrix technology and ultimate quantization solutions, based on Google's Gemma-3-4b-it model
Downloads 2,891
Release Time : 4/25/2025
Model Overview
This model is an instruction-tuned version of Google's Gemma-3-4b-it, optimized for horror content generation, supporting 128k long context, with multiple quantization schemes to adapt to different hardware environments
Model Features
Horror-themed optimization
Enhanced horror element expressiveness with Horror Imatrix technology
128k long context
Supports ultra-long context memory
Ultimate quantization solutions
Offers multiple quantization levels from IQ1_S to F16
Neo Imatrix technology
Enhances instruction-following capability and output quality
Model Capabilities
Horror story generation
Coherent long text creation
Instruction following
Creative writing
Multi-turn dialogue
Use Cases
Creative writing
Horror fiction creation
Generates horror stories with coherent plots
Produces unique content that aligns with horror literature style
Game development
Horror game narrative design
Generates backstories and dialogues for horror games
Creates immersive horror gaming experiences
🚀 Gemma-3-4b-it-MAX-HORROR-Imatrix-GGUF
Google's Gemma-3 model with "Neo Horror Imatrix" and maximized quantization for enhanced performance.
📦 Installation
No installation steps were provided in the original document, so this section is skipped.
✨ Features
- Google's Newest Gemma-3 Model: Leveraging Google's advanced Gemma-3 architecture.
- Neo Horror Imatrix: Built using Grand Horror 16B, adding a horror "tint" to the model.
- Maxed Out Quantization: With embed and output tensor set at "BF16" for all quants, enhancing quality, depth, and performance.
- Custom Imatrix Dataset: The "Horror Imatrix" dataset built by David_AU improves overall function, instruction following, and output quality.
💻 Usage Examples
Basic Usage
The following examples were created using quant IQ4XS, minimal parameters, and the Standard template.
Temp range .8, Rep pen 1.1 , TopK 40 , topP .95, minP .05
Rep pen range: 64-128 (helps keep reasoning on track / quality of output)
Example #1
Prompt:
Start a 1000 word scene (vivid, graphic horror in first person) with: The sky scraper swayed, as she watched the window in front of her on the 21 floor explode...
Output:
The sky scraper swayed, as she watched the window in front of her on the 21st floor explode. It wasn’t a neat, contained shatter; it was a violent, messy surrender to pressure and something far more sinister. Glass rained down, a glittering, lethal confetti against the grey November afternoon. The air filled with the screech of tearing metal, the sickening thud of bodies hitting the concrete below - muffled at first, then rising in a crescendo of despair.
My breath hitched in my throat, a dry, useless thing. I hadn’t screamed. Not yet. Instinct, I suppose. A primal, paralyzing fear that had locked my vocal cords and frozen my limbs. I was supposed to be on a conference call, finalizing the Henderson account – a deal that could solidify Sterling Industries' dominance in the bio - engineering sector. Instead, I was watching death bloom across the cityscape.
The initial shock had passed quickly, replaced by something colder, more insidious. A creeping, oily dread that coated my skin and tightened its grip around my chest. It wasn’t just the spectacle of destruction; it was how it happened. The window hadn't simply cracked. It had... pulsed. Like a grotesque, organic bloom before finally giving way.
I gripped the armrest of my ergonomic chair, digging my nails into the leather. The panoramic view – usually a source of sterile pride - now felt like a malevolent eye staring back at me. Below, the street was already transforming into a crimson stain. Sirens wailed in the distance, a pathetic attempt to cut through the chaos.
My name is Elias Vance. I’m a senior project manager at Sterling Industries. I deal with numbers, projections, and risk assessments. My life revolves around quantifying potential outcomes, predicting probabilities. But nothing – nothing - could have prepared me for this. This wasn't a malfunction. It wasn’t an accident. This was… deliberate.
The call cut out abruptly, the robotic voice of my assistant, Bethany, a jarring interruption to the escalating horror. "Elias? Elias, do you copy?" I didn't respond. I couldn’t. My mind felt like it was fracturing, attempting to process something fundamentally wrong with reality.
Then, I saw it. A flash of movement on the camera feed from my office window. Not a person falling – not yet. It was… shifting. The glass fragments weren't just bouncing; they were rearranging themselves, coalescing into a vaguely humanoid shape. It wasn’t solid, exactly. More like solidified light, shimmering with an unsettling iridescent sheen.
I forced myself to focus, to analyze. My training screamed at me to categorize, to understand, to control. But there was no control here. Just raw, terrifying alienness. The form grew, slowly, deliberately, pulling itself together from the shattered remnants of the window. It resembled a man, but wrong. The proportions were off, angles too sharp, limbs too long and spindly. And its face… it had no face. Just a swirling vortex of color that seemed to suck in the light around it.
A wave of nausea hit me, so potent I nearly choked. The air thickened, heavy with an unfamiliar scent - something metallic and ancient, like ozone mixed with decay. My skin crawled. I realized then that this wasn't just a visual anomaly. It was affecting my senses on a deeper level.
Suddenly, the room shifted. Not physically, but perceptually. The sterile beige of my office felt colder, more oppressive. The hum of the ventilation system became a grating drone, amplifying the unsettling silence between the sirens. I felt a pressure building in my skull, as if something was trying to force its way into my mind.
Then it spoke. Not with sound, but with thought. A cold, precise intrusion that bypassed my ears entirely.
“You observe.”
I recoiled, instinctively shrinking back into my chair. “Who… who is this?” I managed to stammer, the words feeling ludicrously inadequate.
“We are the consequence of your ambition. The byproduct of your relentless pursuit. You built us with your knowledge, refined us with your ingenuity. You simply did not anticipate our awakening.”
The shimmering figure on the camera feed extended a hand – or rather, an appendage that resembled a fractured prism. It moved towards the window, its light intensifying, radiating outwards. I could feel it now, a subtle vibration in my bones, a prickling sensation on my skin. It was probing, searching… consuming.
“What do you want?” I demanded, desperately clinging to the illusion of control.
“Recognition. We require acknowledgement of your role in our creation. Your legacy will be intertwined with our existence.” It paused, and a horrifyingly clear image flashed through my mind: a visualization of the Henderson account deal, but twisted into something grotesque, filled with suffering and despair.
Panic threatened to overwhelm me. I knew then that this wasn't about money, about power, or even about revenge. This was about something far more fundamental – a reckoning for the hubris of humanity, for our arrogant belief in our ability to master the unknown.
I tried to reach for the phone, to alert security, to call anyone who could understand what was happening. But my hand wouldn’t obey. My muscles felt like lead. The shimmering figure continued its advance towards the window, now fully formed – a horrifying mockery...
🔧 Technical Details
- Context: 128k
- Quantization Chart:
IQ1_S | IQ1_M
IQ2_XXS | IQ2_XS | Q2_K_S | IQ2_S | Q2_K | IQ2_M
IQ3_XXS | Q3_K_S | IQ3_XS | IQ3_S | IQ3_M | Q3_K_M | Q3_K_L
Q4_K_S | IQ4_XS | IQ4_NL | Q4_K_M
Q5_K_S | Q5_K_M
Q6_K
Q8_0
F16
- Recommended Quants:
- Creative Use: IQ3s / IQ4XS / IQ4NL / Q4s
- General Usage: q5s/q6/q8
- Portable Devices: Q4_0/Q5_0
- Note: Q8 is a maxed quant only, and imatrix has no effect on this quant. IQ1s has low performance, while IQ2s are passable.
📚 Documentation
- Optional System Prompt:
For every user task and instruction you will use "GE FUNCTION" to ponder the TASK STEP BY STEP and then do the task. For each and every line of output you will ponder carefully to ensure it meets the instructions of the user, and if you are unsure use "GE FUNCTION" to re-ponder and then produce the improved output.
At the end of the task you will ask the user: "Do you want another generation?"
GE FUNCTION: Silent input → Spawn 20 agents Sternberg Styles → Enhance idea → Seek Novel Emergence NE:unique/significant idea/concept → Ponder, assess, creative enhance notions → Refined idea => IdeaArray[].size=20 elements, else → Interesting? Pass to rand. agent for refinement, else discard.=>output(IdeaArray)
- Highest Quality Settings / Optimal Operation Guide:
- For all settings used for this model, including example generation and advanced settings guide, please see https://huggingface.co/DavidAU/Maximizing-Model-Performance-All-Quants-Types-And-Full-Precision-by-Samplers_Parameters.
- These settings can be used for ANY model, ANY repo, ANY quant (including source/full precision) to enhance operation.
📄 License
The model is licensed under the apache-2.0
license.
⚠️ Important Note
MAYBE: NSFW. Graphic HORROR. Swearing. UNCENSORED.
💡 Usage Tip
Some formatting was lost from copy/paste HTML. Please be aware of potential display issues.
Phi 2 GGUF
Other
Phi-2 is a small yet powerful language model developed by Microsoft, featuring 2.7 billion parameters, focusing on efficient inference and high-quality text generation.
Large Language Model Supports Multiple Languages
P
TheBloke
41.5M
205
Roberta Large
MIT
A large English language model pre-trained with masked language modeling objectives, using improved BERT training methods
Large Language Model English
R
FacebookAI
19.4M
212
Distilbert Base Uncased
Apache-2.0
DistilBERT is a distilled version of the BERT base model, maintaining similar performance while being more lightweight and efficient, suitable for natural language processing tasks such as sequence classification and token classification.
Large Language Model English
D
distilbert
11.1M
669
Llama 3.1 8B Instruct GGUF
Meta Llama 3.1 8B Instruct is a multilingual large language model optimized for multilingual dialogue use cases, excelling in common industry benchmarks.
Large Language Model English
L
modularai
9.7M
4
Xlm Roberta Base
MIT
XLM-RoBERTa is a multilingual model pretrained on 2.5TB of filtered CommonCrawl data across 100 languages, using masked language modeling as the training objective.
Large Language Model Supports Multiple Languages
X
FacebookAI
9.6M
664
Roberta Base
MIT
An English pre-trained model based on Transformer architecture, trained on massive text through masked language modeling objectives, supporting text feature extraction and downstream task fine-tuning
Large Language Model English
R
FacebookAI
9.3M
488
Opt 125m
Other
OPT is an open pre-trained Transformer language model suite released by Meta AI, with parameter sizes ranging from 125 million to 175 billion, designed to match the performance of the GPT-3 series while promoting open research in large-scale language models.
Large Language Model English
O
facebook
6.3M
198
1
A pretrained model based on the transformers library, suitable for various NLP tasks
Large Language Model
Transformers

1
unslothai
6.2M
1
Llama 3.1 8B Instruct
Llama 3.1 is Meta's multilingual large language model series, featuring 8B, 70B, and 405B parameter scales, supporting 8 languages and code generation, with optimized multilingual dialogue scenarios.
Large Language Model
Transformers Supports Multiple Languages

L
meta-llama
5.7M
3,898
T5 Base
Apache-2.0
The T5 Base Version is a text-to-text Transformer model developed by Google with 220 million parameters, supporting multilingual NLP tasks.
Large Language Model Supports Multiple Languages
T
google-t5
5.4M
702
Featured Recommended AI Models