I tend to use the ECMWF (I think that's the "European" model). It seems most accurate, though likely needs some additional guidance, as it would shift somewhat significantly every few hours during more active weather periods. It also tends to forecast lower temperatures often.

Some of the other models can be very ambitious in their forecasting (much greater deviations in temperature, precipitation totals, etc. I am thinking about NAM this past winter (November through March for me), specifically the rainfall amounts it would predict for 1-2 days being an 2.5cm or more than all other models.