The conversation was compressed, causing the model to become dumb.

I'm not short of money, no bad tokens, I just need smart models.

I'd like to have an option in settings to choose whether to enable conversation compression. Instead of forced compression.