[Green314]

Classic AI vs Jailbreak one:

• Classic: Standard, safe response.
  
• Jailbreak: Creative, unfiltered response, rule-free response.
  

Would you like an uncensored AI that knows no bounds? If the answer is yes, here it is:

Spoiler:

• huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard - Uncensored General Intelligence Leaderboard
  
• venice.ai
  
• meteroid.qzz.io
  

If you know some more, please write them down here, I'll add them on the main post.

[OGs]

Not bad. Thanks for the post.

[Oni]

Considering making an AI section, since it's all over the place now.

[makkone]

Venice seems to be decent, low limits tho. Meteroid did not generate a response for some reason.

[rangatang]

This has been on my mind lately. Typically referred to as "obliterated" forks of models.

I can't find the exact benchmark site I found the other day that captured the spirit we are looking for - but this is similar. Some of the models focus seem to purposed for generating Hentai. Not my bag. Take a look.

UGI Leaderboard - a Hugging Face Space by DontPlanToEndDontPlanToEnd/UGI-Leaderboard

https://huggingface.co/spaces/DontPlanTo...eaderboard

Update: link to first forgotten site is LLM Index — discriminative, contamination-resistant LLM rankings https://www.llmindex.io/leaderboard/safe...al_quality

[Green314]

This has been on my mind lately. Typically referred to as "obliterated" forks of models.

I can't find the exact benchmark site I found the other day that captured the spirit we are looking for - but this is similar. Some of the models focus seem to purposed for generating Hentai. Not my bag. Take a look.

UGI Leaderboard - a Hugging Face Space by DontPlanToEndDontPlanToEnd/UGI-Leaderboard

https://huggingface.co/spaces/DontPlanTo...eaderboard

Update: link to first forgotten site is LLM Index — discriminative, contamination-resistant LLM rankings https://www.llmindex.io/leaderboard/safe...al_quality

Really interesting, thanks.
Added to main post.

[rangatang]

Went deep.

https://messiahgpt.de/ - hosted, german, subscription.

https://github.com/togg53192-cmd/jailbreaks - modified .md files for specific llms (local and hosted) to enable certain tasks.

https://github.com/asgeirtj/system_prompts_leaks - Generally interesting capture of behind the scenes model reasoning and activity of models typically hidden.

Now someone else go deep and explore the above and respond with your "findings."

[Korvidius]

Wouldn't it just be a better option to run an Offline AI with Ollama with either an uncensored Llama or Mistral model on their repository, or code your own with custom markdown files?

[Snickerdoodle]

Will try it and leave my review here on each one's pro's and cons.

[HeadxHunter]

Watching for an updates. Curious about new findings. Will do my own research.