Since I can't get a clear answer about how to use token-based condensing with the GUI: What is the appropriate memory condenser max history size for models loaded with 64k context size?

This might be a better question for #C06R25BT5B2