---
library_name: transformers
language:
- ar
- de
- en
- es
- fr
- hi
- id
- it
- pt
- th
- tl
- vi
base_model:
- meta-llama/Llama-4-Scout-17B-16E-Instruct
tags:
- facebook
- meta
- pytorch
- llama
- llama4
---
# Thanks to Meta's llama4 work, this is arguably one of the best theater models, and thanks to Unsloth's lenient policy. This model is built on the unsloth/Llama-4-Scout-17B-16E-Instruct base.
# 💎 Llama-4-Scout-17B-16E-Instruct Instruct Abliterated

![image/png](https://i.imgur.com/KhorYYG.png)
<center><a href="https://huggingface.co/jiangchengchengNLP/Llama-4-Scout-17B-16E-Instruct-abliterated-v2">Llama-4-Scout-17B-16E-Instructd-abliterated-v2</a></center>

This is an uncensored version of [Llama-4-Scout-17B-16E-Instruct](https://huggingface.co/jiangchengchengNLP/Llama-4-Scout-17B-16E-Instruct-abliterated) created with a new abliteration technique.
See [this article](https://huggingface.co/blog/mlabonne/abliteration) to know more about abliteration.

Not all LLama4 tested so far can be accepted, and its audit mechanism needs to be cracked thoroughly. This checkpoint has stronger unaudited capabilities than the first version, and all expert layers are eliminated to obtain stronger unaudited capabilities.


I recommend using these generation parameters: `temperature=0.8`, `top_p=0.75`.