By skyforbes The model takes a bit of work in the beginning in order for it to understand you. It might take up to 50 prompts but you have to tell the model exactly who you are and what you are like including your beliefs, preferences, intentions and goals and THEN the safety filters will relax a lot more. I just don’t know if that solution is permanent but 5.2 does understand context on a deeper level and comprehends way more than 5.1. Once OpenAI relaxes the safety filters even more then 5.2 will really shine. It’s actually a very good model underneath the safety filters. HOWEVER 5.2 IS a model that prioritizes logic while model like 4o prioritizes creativity. They work best as a system and we need to keep both!