groxaxo/Huihui-gemma-4-26B-A4B-it-abliterated-GGUF

Huihui Gemma 4 26B A4B IT Abliterated — GGUF Quantizations

Overview

Huihui-gemma-4-26B-A4B-it-abliterated-GGUF is a GGUF release for llama.cpp-compatible runtimes and local inference, published by groxaxo. It is intended for open-source evaluation, reproducible experimentation, and compatible local or hosted inference workflows. The wording below is deliberately limited to what can be verified from this repository's metadata and artifacts.

The repository name identifies a behavior-modified or reduced-filtering lineage. That label describes the source or conversion history; it is not a guarantee of unrestricted behavior in every prompt or runtime. Test outputs carefully before sharing or deploying them.

At a glance

Field	Details
Format	GGUF
Source / base	huihui-ai/Huihui-gemma-4-26B-A4B-it-abliterated
Intended task	image-text-to-text
License	the license declared in the repository files

What is included

• *.gguf (22 files)
• config.json
• generation_config.json
• tokenizer.json
• tokenizer_config.json
• processor_config.json
• chat_template.jinja
• Additional configuration, tokenizer, processor, or shard files (32 visible artifacts total)

Quick start

llama.cpp

Download a .gguf file that fits your available memory, then run it with a current llama.cpp build:

llama-cli \
  -m /path/to/model.gguf \
  -p "Write a concise technical summary."

For vision or any-to-any models, download the matching multimodal projection file when one is provided and follow the source model's modality-specific instructions.

Compatibility and responsible use

• Use a runtime that explicitly supports this format, architecture, and modality.
• Keep configuration, tokenizer, processor, projection, and weight files from the same revision together.
• Review the source model card and license before redistribution or deployment.
• Hardware needs depend on parameter count, context length, cache precision, quantization, and concurrency.
• Report reproducible issues with the runtime version, hardware, launch command, and a minimal example.

Quantization or conversion changes numerical behavior, memory use, and throughput relative to the source checkpoint; validate quality on your own workload.

Generated outputs may be inaccurate or unsuitable for a given use case. Users are responsible for testing behavior, applying appropriate safeguards, and complying with applicable licenses and laws.

This repository contains GGUF / llama.cpp quantized builds of:

huihui-ai/Huihui-gemma-4-26B-A4B-it-abliterated

These are UD quantizations prepared for efficient local inference with llama.cpp, including support for multimodal image-text-to-text workflows when used with the corresponding mmproj file.

Overview

This release is designed for users who want to run the Huihui Gemma 4 26B A4B abliterated model locally with reduced VRAM and RAM requirements while preserving as much output quality as possible.

The quantization variants use an optimized tensor distribution strategy inspired by Unsloth-style mixed-quality quantization recipes, balancing model fidelity, speed, and memory efficiency across different hardware targets.

Quick Start

1. Download the latest release of llama.cpp.
2. Download your preferred .gguf model file from this repository.
3. For multimodal inference, also download the matching mmproj file.
4. Run the model with llama.cpp using your preferred frontend or CLI.

Example:

./llama-cli \
  -m Huihui-Gemma-4-26B-A4B-it-abliterated-UD-Q4_K_XL.gguf \
  --mmproj mmproj-model.gguf \
  -p "Describe this image in detail."

Adjust the model filename and mmproj filename to match the files you downloaded.

Which Quant Should I Choose?

Choose based on your available memory and quality target:

• Higher-bit / larger quants: Better quality, higher VRAM/RAM usage.
• Mid-range quants: Best balance for most local setups.
• Lower-bit quants: Faster and smaller, but with more quality loss.

For best results, use the largest quantization your hardware can comfortably run.

Multimodal Usage

This model supports image-text-to-text inference when used with the appropriate multimodal projection file.

Make sure the mmproj file matches this model family. Using an incorrect projection file may result in broken or degraded vision-language behavior.

Notes

• This is a quantized GGUF release of the fine-tuned model.
• Original model: huihui-ai/Huihui-gemma-4-26B-A4B-it-abliterated
• Runtime target: llama.cpp
• Format: GGUF
• Modality: image-text-to-text
• Quantization style: UD / mixed tensor distribution

Disclaimer

This repository only provides quantized GGUF builds. Model behavior, alignment characteristics, and training details are inherited from the original base model and fine-tune.