Sigmund Fraud
893 all the way.
.
Square Leg, the table-top exercise that produced the scenario considered most likely in the early 80s, and which Threads was based on, assumed 200 or thereabouts. Still enough to obliterate most built-up areas and kill half the population at a stroke, though.
The snappily titled Greater London Area War Risk Study (GLAWARS) from 1986 examines two main attack scenarios, one with 9MT hitting the capital and one with 12MT. The chiling conclusion (and I stupidly read the whole book of 20 odd separate studies with a terrible flu on the go) was that anything more than 9MT was unnecessary; 9MT was enough to destroy 97% of all housing stock in all London boroughs and kill 70% of the population instantly with the survival rate dropping down to just tens of thousands after attack +5days. 12MT just made more ash.