AIbase
EN
Home
Gemma 3 4b It Abliterated GGUF
G

Gemma 3 4b It Abliterated GGUF

Developed by ZeroWw
An innovative quantization solution that achieves smaller model size while maintaining high performance through mixed-precision quantization.
Large Language Model EnglishOpen Source License:MIT#Mixed-precision quantization#Efficient text generation#f16 embedding optimization
Downloads 247
Release Time : 3/22/2025

Model Overview

This model adopts a self-developed mixed-precision quantization scheme, using f16 precision for the output and embedding layers and q5_k or q6_k precision for the remaining parts, achieving a smaller size than standard q8_0 while maintaining performance comparable to pure f16 quantization.

Model Features

Mixed-precision quantization
Uses f16 precision for the output and embedding layers and q5_k or q6_k precision for the remaining parts, achieving efficient quantization.
Size optimization
Both f16.q6 and f16.q5 quantization schemes result in a smaller size than standard q8_0 quantization.
Performance retention
Quantized performance remains on par with pure f16 quantization.

Model Capabilities

Text generation

Use Cases

Natural language processing
Efficient text generation
Reduces model size while maintaining generation quality.
Smaller size than standard q8_0 with performance comparable to pure f16 quantization.
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
English简体中文繁體中文にほんご
© 2025AIbase