I'm using pyomo 6.8.2 and some open source solvers downloaded recently: CBC, SCIP, HIGHS, MBNB

I have a not so big MILP problem (large python program that cannot be shown where ) with about a dozen of binaries and a hand full of integers. I generate about 25000 configurations in one run and solve these step by step.

Using python-mip with

cbc takes on average 0.04 sec (internally)
have to remove becaus call does not work properly and cbc was used _scip takes on average 0.04 sec (from path)_
have to remove becaus call does not work properly and cbc was used _highs takes on average 0.04 sec (from path)_

Using HiGHS with highspy takes about 0.057 sec.
Using SCIP with pyScipOpt takes about 0.04 sec.

Using Pyomo with

cbc takes on average 0.14 sec (from path)
scip takes on average 0.13 sec (from path)
highs takes on average 0.13 sec (from path)

so there seems to be a "significant" overhead of 0.1 sec for each run.

I redirected the tempdir to a ramdisc to avoid disc overhead but this doesn't seem to help.

Is there any chance to speed up this binding overhead?