Tetratype/Gemma-4-26B-A4B-it-heretic-antislop

VISIONConcurrency Cost:2Model Size:26BQuant:FP8Ctx Length:32kTool Calling:SupportedPublished:May 26, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

Tetratype/Gemma-4-26B-A4B-it-heretic-antislop is a 26 billion parameter language model, based on the Gemma-4 architecture, that has been fine-tuned using Sam Paech's auto-antislop pipeline. This model is specifically designed to mitigate undesirable outputs, making it suitable for applications requiring robust content moderation and adherence to specific safety guidelines. With a context length of 32768 tokens, it offers extensive input processing capabilities for various text-based tasks.

Loading preview...

Model Overview

Tetratype/Gemma-4-26B-A4B-it-heretic-antislop is a 26 billion parameter language model built upon the Gemma-4 architecture. This model distinguishes itself through its specialized fine-tuning process, which leverages Sam Paech's auto-antislop pipeline.

Key Capabilities

  • Content Moderation: The primary focus of this model's fine-tuning is to reduce or eliminate undesirable outputs, making it inherently more aligned with specific content guidelines.
  • Robustness: The 'antislop' methodology aims to enhance the model's reliability in generating appropriate responses across a wide range of prompts.
  • Extended Context: With a context window of 32768 tokens, it can process and understand longer inputs, facilitating more complex interactions and document analysis.

Good For

  • Applications requiring strict content control: Ideal for chatbots, virtual assistants, or content generation systems where output safety and adherence to guidelines are paramount.
  • Mitigating harmful or off-topic responses: Useful in scenarios where preventing 'slop' or irrelevant/inappropriate content is a critical requirement.
  • Developers seeking a more controlled Gemma-4 variant: Offers a pre-tuned solution for those who need a Gemma-4 model with enhanced output filtering capabilities.