geekcy2333/gemma-4-26B-A4B-it-ultra-uncensored-heretic

VISIONConcurrency Cost:2Model Size:26BQuant:FP8Ctx Length:32kTool Calling:SupportedPublished:Apr 7, 2026License:apache-2.0Architecture:Transformer Open Weights Cold

geekcy2333/gemma-4-26B-A4B-it-ultra-uncensored-heretic is a 26 billion parameter instruction-tuned Gemma 4 model, developed by geekcy2333, that has been decensored using the Heretic v1.2.0 Arbitrary-Rank Ablation (ARA) method. This model significantly reduces content refusals (3/100 vs 100/100 for the original) while largely preserving the original model's quality, as indicated by a low KL divergence of 0.1237. It is designed for applications requiring less restrictive content generation, maintaining strong performance in reasoning and multimodal tasks.

Loading preview...

Model Overview

This model, geekcy2333/gemma-4-26B-A4B-it-ultra-uncensored-heretic, is a decensored version of Google's Gemma-4-26B-A4B-it, created using the Heretic v1.2.0 tool with the Arbitrary-Rank Ablation (ARA) method. Its primary differentiator is a drastic reduction in content refusals, achieving 3/100 refusals compared to the original model's 100/100, while maintaining a low KL divergence of 0.1237, indicating strong preservation of the original model's quality.

Key Capabilities

  • Reduced Refusals: Significantly fewer content restrictions compared to the base model.
  • Multimodal: Inherits Gemma 4's capabilities for text and image input, with text output.
  • Reasoning: Designed for strong reasoning tasks, with configurable thinking modes.
  • Long Context: Supports a 256K token context window.
  • Coding & Agentic Capabilities: Enhanced performance in coding benchmarks and native function-calling support.

Good For

  • Applications requiring less restrictive content generation.
  • Tasks benefiting from multimodal understanding (text and image).
  • Reasoning, coding, and agentic workflows where the base Gemma 4 model's capabilities are desired without its inherent content restrictions.

Finetunes
1 models