If you want 12 digit resolution, or to be more precise 10-12(what's your application, anyhow?), that's not easily done with any counter in 1sec.
The internal time interpolation method usually determines the 1sec resolution, so this should be 100ps or 10-10/sec for the '132, but 500ps or 5*10-10 only for the '131.
Both counters use sort of running average, which virtually increases the resolution/sec.
12digits/sec I think is not a reasonable specification.
Somewhere on time-nuts, it's described that this mechanism falsifies somehow the allan distribution statistics, and if you want to calibrate or compare e.g. a Rb vs. GPSDO, you should use different methods than frequency measurement, to achieve real 10-12 resolution/accuracy.
This method would better be a phase detection, or T.I. measurement between both sources, accumulated over a longer period of time, like 1000...10000sec, using TimeLab or similar.
For that purpose, even 1ns is enough resolution, but 100ps would give faster results.
Btw. the old 5370B has 20ps T.I. resolution and yields real 12 digits or 10-11 in one sec.
Frank