Performance time

[mastercracker3d]

Hi guys, i work in a fork of GPT4ALL (if any component will be compatible with the definitions of the community i will to provide).
I executed differents models and all get the response in one minute or almost in 3.
My fork execute questions in parallel (patterns Master-worker and Processing Pattern of XTP).
if Anybody obtain better results of the GPT4ALL by example with a GPU Model XXX seconds or milliseconds for the response, please if is possible to obtain the information for to dimensioning a groups of machines?