# Jail breaking open models — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/a3a9fbc99ba90527

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-09-27 01:07 ingested 2026-09-28 02:41

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1wr6u8m/jail_breaking_open_models/)

## retained text



```
Is there any resource dedicated to jail breaking open models?  reddit, discord, etc?  I know some of the models yield easily, but some of them can be stubborn especially the large smarter ones.   I have tried uncensored models and while they might pass sometimes, they often end up doing stupid things the censored ones don't.  No matter the claims, it seems altering the weights ends up affecting the intelligence.    

If anyone knows any techniques, please share or point me towards the right resources.
```



document fetched 2026-09-28 00:39 · canonical https://reddit.com/r/LocalLLaMA/comments/1wr6u8m/jail_breaking_open_models
