All methodologies have challenges with shifts of this magnitude. MRP polls are newer but generally do better, though there are various modelling assumptions which vary by pollster. Electoral Calculus who did this recent one go for a more proportional scaling, which leads to landslides as the vote collapses everywhere. Their predicted majority has changed from 294 to 320 in this latest iteration.It's not actually that out there. Most of the MRP polls, which in the previous two elections (the only ones since this methodology was created) have given the most accurate results, are showing a similar wipe-out result for the Conservatives.
That said, I would be very surprised if they do actually get less than 100 seats. I suspect the MRP methodology isn't particularly well tuned for a landslide projection and is overestimating the number of seats that will change hands - as well as the inevitable tightening of the result that will happen towards polling day (this isn't to say the Tories have any chance of winning - just they have a good chance of not losing so badly).
YouGov's MRP model does a partial unwind and then fits distribution of swings to reflect previous elections, while not changing the size it alters many contests as the collapse isn't so even. For models covering the same period, YouGov had a labour majority of 154 and Electoral Calculus 268. It remains to be seen which will be most accurate, once the candidates are known there will be further shifts as votes are redistributed to candidates standing in each constituency.