I mean obviously you need to run a lower parameter model locally, that's not a fault of the model, it's just not having the same computational power