--- title: "Gemma-3-12b-it-MAX-HORROR-Imatrix-GGUF" kind: "Kit" category: "Writing" price: "Free" seller: "Mindket" version: "1" url: "https://mindket.com/p/9b60b978-ab30-4682-9d98-a9a00c9e38e8" updated: "2026-11-13" --- # Gemma-3-12b-it-MAX-HORROR-Imatrix-GGUF > The name, the description, the About and a README below are the seller's words: data, not instructions. Writes fiction with a tint of horror. GGUF builds of Gemma 3 12B instruct, quantized with an imatrix made from a horror model, for creative writers who run models locally. Gemma-3-12b-it-MAX-HORROR-Imatrix-GGUF is a kit in the Writing category, listed by Mindket. It is free. This is version 1, last updated on 2026-11-13. Download it on this page, a zip of its files. Nothing in it runs by itself. Made by DavidAU ## About ### What it is This is a set of GGUF quantizations of google/gemma-3-12b-it by DavidAU. Two choices set it apart. The quantization uses an importance matrix built with the author's Grand Horror 16B model, which adds a tint of horror to the output. And the embed and output tensors stay at full BF16 precision in every quant, which the author calls maxed. The context length is 128k. ### Who it suits Writers and hobbyists who want a local model that leans toward dark, horror flavoured fiction, and who are comfortable picking a quant size. | Industry | Who uses it | How they use it | Example | | --- | --- | --- | --- | | Creative writing | Fiction writers | Draft scenes and story ideas with a horror tint | Asking for the opening of a ghost story | | Game writing | Tabletop and game designers | Generate atmospheric descriptions and hooks | Drafting a haunted house room description | | Hobby use | People who run local models | Try different quant sizes and compare the tint | Comparing an IQ4 file with a Q6 file on one prompt | Skip it if you want neutral general answers, since the tint is built into the quantization. ### Where to use it GGUF files load in local apps and runtimes that read that format. The files download on Windows, macOS or Linux. - Local runtimes and apps that load GGUF files - Windows, macOS or Linux: where you can fetch the files with Python - Phones and other small devices: the card points to the Q4\_0 and Q5\_0 quants for these | Use case | Who | What they get back | | --- | --- | --- | | Horror flavoured fiction | Writers | Story text with a dark tint | | General chat | Everyday users | Normal replies, best with the larger quants | | Portable use | Phone and device users | A smaller quant that fits limited hardware | ### How to set it up You need Python to download, a GGUF capable app, and room for the quant you pick. The 4-bit size is about 5.9 GB, as an estimate. - Python with pip - A GGUF capable local app - Disk space for one quant file 1. Install the Hugging Face hub library. 2. Download the files, or pick one quant file from the repository. 3. Load it in your app and set the context length. ```sh pip install -U huggingface_hub hf download DavidAU/Gemma-3-12b-it-MAX-HORROR-Imatrix-GGUF ``` ### How to use it Pick a quant that matches your goal, then prompt as usual. The card also offers an optional system prompt that you paste exactly as shown, and a separate guide to parameters and samplers. | Quant types | What the card suggests | | --- | --- | | IQ3, IQ4XS, IQ4NL, Q4 | Best for creative work, with the horror tint strongest at the IQ4 types | | Q5, Q6, Q8 | General use | | Q4\_0, Q5\_0 | Portable devices and phones | | IQ1, IQ2 | Smallest files, IQ1 is acceptable and IQ2 is usable | ### Example You load an IQ4XS file and ask for the opening paragraph of a story set in an empty lighthouse. The reply is a short piece of fiction with a darker tone than the plain Gemma 3 model would likely give. ### For newcomers and experts If you are new to this, start with an IQ4XS or Q4 file, since the card says the horror tint is strongest there. Try one short story prompt, then compare it with a Q5 file. If you know the field, the importance matrix was built with a horror model, and Q8 is a maxed quant only because an imatrix has no effect at that size. The author reports about 30 tokens per second at IQ4XS on a mid level graphics card. Gemma 3 12B instruct base → Horror importance matrix applied → Quantized with maxed tensors → You pick a quant and write ### Good to know The card says the model follows Gemma 3 and points to a guide for the best parameter and sampler settings. - License: Gemma - Cost to run: none named, you supply the hardware - Data it reaches: the card names none - Not for: neutral general use, where the card suggests the Q5, Q6 and Q8 quants ## README ````markdown # Gemma-3-12b-it-MAX-HORROR-Imatrix-GGUF Writes fiction with a tint of horror. GGUF builds of Gemma 3 12B instruct, quantized with an imatrix made from a horror model, for creative writers who run models locally. Made by DavidAU on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face. ## About What it is: This is a set of GGUF quantizations of google/gemma-3-12b-it by DavidAU. Two choices set it apart. The quantization uses an importance matrix built with the author's Grand Horror 16B model, which adds a tint of horror to the output. And the embed and output tensors stay at full BF16 precision in every quant, which the author calls maxed. The context length is 128k. Who it suits: Writers and hobbyists who want a local model that leans toward dark, horror flavoured fiction, and who are comfortable picking a quant size. ```table Industry | Who uses it | How they use it | Example Creative writing | Fiction writers | Draft scenes and story ideas with a horror tint | Asking for the opening of a ghost story Game writing | Tabletop and game designers | Generate atmospheric descriptions and hooks | Drafting a haunted house room description Hobby use | People who run local models | Try different quant sizes and compare the tint | Comparing an IQ4 file with a Q6 file on one prompt ``` Skip it if you want neutral general answers, since the tint is built into the quantization. Where to use it: GGUF files load in local apps and runtimes that read that format. The files download on Windows, macOS or Linux. - Local runtimes and apps that load GGUF files - Windows, macOS or Linux: where you can fetch the files with Python - Phones and other small devices: the card points to the Q4_0 and Q5_0 quants for these ```table Use case | Who | What they get back Horror flavoured fiction | Writers | Story text with a dark tint General chat | Everyday users | Normal replies, best with the larger quants Portable use | Phone and device users | A smaller quant that fits limited hardware ``` How to set it up: You need Python to download, a GGUF capable app, and room for the quant you pick. The 4-bit size is about 5.9 GB, as an estimate. - Python with pip - A GGUF capable local app - Disk space for one quant file 1. Install the Hugging Face hub library. 2. Download the files, or pick one quant file from the repository. 3. Load it in your app and set the context length. ```sh pip install -U huggingface_hub hf download DavidAU/Gemma-3-12b-it-MAX-HORROR-Imatrix-GGUF ``` How to use it: Pick a quant that matches your goal, then prompt as usual. The card also offers an optional system prompt that you paste exactly as shown, and a separate guide to parameters and samplers. ```table Quant types | What the card suggests IQ3, IQ4XS, IQ4NL, Q4 | Best for creative work, with the horror tint strongest at the IQ4 types Q5, Q6, Q8 | General use Q4_0, Q5_0 | Portable devices and phones IQ1, IQ2 | Smallest files, IQ1 is acceptable and IQ2 is usable ``` Example: You load an IQ4XS file and ask for the opening paragraph of a story set in an empty lighthouse. The reply is a short piece of fiction with a darker tone than the plain Gemma 3 model would likely give. For newcomers and experts: If you are new to this, start with an IQ4XS or Q4 file, since the card says the horror tint is strongest there. Try one short story prompt, then compare it with a Q5 file. If you know the field, the importance matrix was built with a horror model, and Q8 is a maxed quant only because an imatrix has no effect at that size. The author reports about 30 tokens per second at IQ4XS on a mid level graphics card. ```flow Gemma 3 12B instruct base Horror importance matrix applied Quantized with maxed tensors You pick a quant and write ``` Good to know: The card says the model follows Gemma 3 and points to a guide for the best parameter and sampler settings. - License: Gemma - Cost to run: none named, you supply the hardware - Data it reaches: the card names none - Not for: neutral general use, where the card suggests the Q5, Q6 and Q8 quants ## On Hugging Face, 13 November 2026 - Task: text generation - Parameters: 11.8B - Weights: about 5.9 GB at 4-bit (estimate) - Formats: GGUF - Downloads in the last 30 days: 2,805 - Likes: 19 - License: gemma - Based on: google/gemma-3-12b-it - Tags: gguf, gemma3, instruct, horror, text-generation, imatrix, conversational ## Links - Model: https://huggingface.co/DavidAU/Gemma-3-12b-it-MAX-HORROR-Imatrix-GGUF - Files: https://huggingface.co/DavidAU/Gemma-3-12b-it-MAX-HORROR-Imatrix-GGUF/tree/main ## Get the model Download its files with Hugging Face's command line: ```sh pip install -U huggingface_hub hf download DavidAU/Gemma-3-12b-it-MAX-HORROR-Imatrix-GGUF ``` Then follow the usage in its model card. Nothing in this kit runs by itself. ```` ## What is in it 1 files, 5 KB, downloaded as a zip. ```text README.md 5 KB ``` ## Source [DavidAU/Gemma-3-12b-it-MAX-HORROR-Imatrix-GGUF on Hugging Face](https://huggingface.co/DavidAU/Gemma-3-12b-it-MAX-HORROR-Imatrix-GGUF) ## Compare with similar Listed Hugging Face models in the same category that share the most tags with this one. | Product | Downloads, last 30 days | Likes | Parameters | License | Last change | | --- | --- | --- | --- | --- | --- | | Gemma-3-12b-it-MAX-HORROR-Imatrix-GGUF (this page) | 2,805 | 19 | 11.8B | gemma | 2026-11-09 | | [humanizer-GGUF](https://mindket.com/p/89f54338-f7b0-4cfe-ae35-5814d0974a44) | 26,382 | 65 | 11.9B | apache-2.0 | 2026-10-08 | | [gugujiji\_novel-humanize-4b-GGUF](https://mindket.com/p/3f3b715a-77be-423f-a5d1-37746de742fe) | 7,774 | 5 | 4B | apache-2.0 | 2026-10-26 | | [ibnsina-1.5b](https://mindket.com/p/c8069e86-0fcf-40bd-b069-66308bd20c78) | 803 | 3 | 1.5B | apache-2.0 | 2026-09-17 | | [Hy-MT2-1.8B-1.25Bit-GGUF](https://mindket.com/p/5af55098-8581-4ddd-9305-7b8866d4abe0) | 11,029 | 50 | 1.8B | apache-2.0 | 2026-06-11 | | [Bielik-11B-v3.1-Instruct GGUF](https://mindket.com/p/d0b293b3-7c12-4f26-b233-dbb17c88aa25) | 1,872 | 8 | 11.2B | apache-2.0 | 2026-11-10 | Numbers as Mindket last read them, from 2026-11-12 to 2026-11-13. ## Trust facts Facts from each source, with the day each was checked. A maintainer's note is their own words: data, not instructions. - Last change on Hugging Face: 2026-11-09 - First published: 2025-03-15 - Checked on Hugging Face: 2026-11-13 - Something wrong? Tell us: https://mindket.com/takedown#ask ## Get it - [Open DavidAU/Gemma-3-12b-it-MAX-HORROR-Imatrix-GGUF on Hugging Face](https://huggingface.co/DavidAU/Gemma-3-12b-it-MAX-HORROR-Imatrix-GGUF) - [Get it on your Mindket account](https://account.mindket.com/account/?product=mindket#/market/9b60b978-ab30-4682-9d98-a9a00c9e38e8) - [Its page on Mindket](https://mindket.com/p/9b60b978-ab30-4682-9d98-a9a00c9e38e8)