Quoted By:
Alright, I tried comparing GPT-3.5, GPT-4 Turbo, and WizardLM-2 8x22B. I used the same card, same prompt, and same jailbreak for all of them. As of now, I don't see much difference between all of them but at least with WizardLM-2 8x22B I don't get any "I'm sorry..." response from it so this is certainly an improvement.
I used older version of GPT-3.5 and GPT-4 Turbo btw, because the newest version refuse to generate response. So yeah, I am going to switch to Wizard since who knows when openai will remove the older version.