For my microclimate, I’ve found that Meteoblue is the only forecast that doesn’t forecast the high temperature 10 degrees too low when there is a heat wave.
That being said, you shouldn't treat these outputs as gospel. All these models will get the big picture correct, but they have limitations with respect to spatial resolution. So if you click on the Windy map and look at your local forecast, take it with a grain of salt - the model doesn't resolve down to the scale of your backyard. The status quo is for companies or agencies to use statistics or ML to "correct" the forecast against local observations, which dramatically improves the accuracy of forecasts like, "What will tomorrow's max temperature be" or, "what is the probability it will rain this evening?"