# Inference time abliteration

**URL:** </static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Finference-time-abliteration%2F190909%3E%5C
**Category:** Show and Tell\
**Created:** [October 22, 2026, 2:48pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Finference-time-abliteration%2F190909 "2026-10-22T14:48:25Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![heterodoxin](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/heterodoxin/32/64521_2.png) [@heterodoxin](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fheterodoxin)%5C
**Post date:** [October 22, 2026, 2:48pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Finference-time-abliteration%2F190909%2F1 "2026-10-22T14:48:25Z")

</div>

AbliteratedOutTheBox is a llama.cpp fork that abliterates any model you load, live at inference, with no fine-tuning and no extra files.

This eliminates any hassle of abliteration without requiring any more vram.

> **[GitHub - heterodoxin/AbliteratedOutTheBox: llama.cpp fork that abliterates at inference](https://github.com/heterodoxin/AbliteratedOutTheBox)**
>
> llama.cpp fork that abliterates at inference