@User, that substitution by constant number was done as I observed that the behavioral models were failing to parse tc1 and tc2 in every case except when they were constant numbers. Even setting something as simple as "tc1 = {tc1rpolybody * 1}" would leave them zeroed out. Thank you for bringing this up. It should have been noted in the issue/patch. I was working with ngspice-33. Could you check if that resistor model shows the same temperature dependence after the changes?