llmfan46/gemma-4-Ortenzya-The-Creative-Wordsmith-31B-it-uncensored-heretic

VISIONConcurrency Cost:2Model Size:31BQuant:FP8Ctx Length:32kTool Calling:SupportedPublished:May 12, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

llmfan46/gemma-4-Ortenzya-The-Creative-Wordsmith-31B-it-uncensored-heretic is a 31 billion parameter instruction-tuned Gemma 4 model developed by llmfan46. It is a finetuned version of a decensored Google Gemma 4 31B model, specifically optimized for creative writing, role-playing, and translations. This model significantly reduces content refusals compared to the original, aiming for a more human-like and less mechanical writing style.

Loading preview...

Model Overview

This model, developed by llmfan46, is a 31 billion parameter instruction-tuned Gemma 4 variant. It was created by first decensoring the original google/gemma-4-31B-it model using the Heretic v1.2.0 tool with Arbitrary-Rank Ablation (ARA) and then finetuned with Unsloth Studio v0.1.39-beta.

Key Differentiators

  • Reduced Refusals: Achieves 9/100 refusals compared to 99/100 in the original gemma-4-31B-it, indicating significantly fewer content restrictions.
  • Enhanced Creative Writing: Specifically finetuned to improve creative writing, role-playing, and translation capabilities, aiming for a more natural, human-like tone over a corporate or robotic style.
  • Uncensored Base: The finetuning process was applied to an already uncensored version of the base model.

Good For

  • Creative Writing: Generating stories, scripts, and other creative content with improved naturalness.
  • Role-Playing: Engaging in more dynamic and less restricted conversational role-play scenarios.
  • Translations: Producing translations that sound less "machine written" and more fluid.
  • Applications requiring reduced content moderation: Use cases where a lower refusal rate is critical for generating diverse content.