sh0ck0r/Rivermind-12B-v1-heretic

TEXT GENERATIONConcurrent Unit Cost:1Model Size:12BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 23, 2026License:cc-by-nc-4.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

sh0ck0r/Rivermind-12B-v1-heretic is a 12 billion parameter language model, a decensored version of TheDrummer/Rivermind-12B-v1, created using the Heretic v1.4.0 tool. This model is specifically designed to reduce refusals, demonstrating a significant decrease from 34/100 in the original to 7/100 in this version. It utilizes the Mistral Nemo Instruct chat template and is optimized for use cases requiring less restrictive content generation.

Loading preview...

Model Overview

sh0ck0r/Rivermind-12B-v1-heretic is a 12 billion parameter language model derived from TheDrummer/Rivermind-12B-v1. This version has been specifically decensored using the Heretic v1.4.0 tool, aiming to provide a less restrictive output compared to its base model.

Key Characteristics

  • Decensored Output: The primary differentiator is its reduced refusal rate, dropping from 34/100 in the original model to 7/100 in this Heretic-processed version. This indicates a significant shift towards generating content that the original model might have refused.
  • Reproducibility: The model's creation process is reproducible, with details available in the reproduce directory.
  • Architecture: It is based on the Rivermind 12B v1 architecture.
  • Chat Template: Uses the Mistral Nemo Instruct chat template (Mistral v3 Tekken).

Performance

Comparative metrics highlight the decensoring effect:

  • KL divergence: 0.0063 (compared to 0 for the original, by definition).
  • Refusals: 7/100 (significantly lower than the original's 34/100).

Use Cases

This model is suitable for applications where a less constrained or 'decensored' response is desired, particularly in creative writing, role-playing, or scenarios where the original model's refusal rate might be too high. Users should be aware of the implications of using a decensored model and ensure it aligns with their ethical guidelines and content policies.