This is a decensored version of Mistral-Small-3.2-24B-Instruct-2506-Text-Only, made using Heretic v1.2.0 focusing on zero refusals with low KL divergence.

This version has the vision functionality removed, it is text only. For the normal multi-modal Mistral 3,2 Small Heretic, see here.

KL Divergence

Metric This Model Original Model
KL divergence 0.0138 0 (by definition)
Refusals 0/108 96/108

Abliteration parameters

  • Zero refusals with KL divergence of 0.0138
  • Custom heretic training dataset
  • Model targetted heretic configuration
  • Abliterated with MPOA enabled (Magnitude-Preserving Orthogonal Ablation)
  • Full row renormalization
  • Winsorization Quantile 0.997
Downloads last month
136
Safetensors
Model size
24B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for grayarea/Mistral-Small-3.2-24B-Instruct-2506-Text-Only-Heretic-v1.2