Hello,
the question is well phrased, yet hard to answer.
The default car following model 'Krauss' was written with the intention of creating a simple model that recreates macroscopic traffic properties (fundamental diagram). As far as I know distinctions between human/autonomous drivers were out of the scope at that time.
Other models such as the Wiedemann model were intended to capture perception effects of human drivers.
Generally, I would say the models were not built with the distinction human/autonomous in mind. Instead they were intended to capture qualitative effects of traffic (why do cars change their lane?, what is a safe following distance) and to allow for flexible calibration.
There are some developments in Sumo such as the driver-state device which are intended to make the driving behavior more human. However, this part of the development is still ongoing (
https://sumo.dlr.de/wiki/Driver_State).
Most of the models allow a wide range of different behaviors which can be used to model various assumptions on how humans and autonomous vehicles might differ (preferred headway, reaction time, rule compliance, imperfection).
regards,
Jakob