[mjsweet]

Could this be related to their plans for local inference? Maybe the need to load larger models into memory has driven their decision?

[bearjaws]

Has to be, even a small model is 4gb of RAM. Open up any O365 based product in Chrome and kiss another 2gb goodbye if you only have 8gb you're left with very little after the OS...

[surfingdino]

So no actual increase in available RAM and no improvement in performance then. That'll be an extra £999, sir.