Re: [PD] GEM - multimodel performance issue