@User: Yes, that should be correct. I don't know why the ngspice simulation is coming up with such a low value. The "rsheet" value in the base model is 2000, and there appears to be no second-order adjustments to the resistor equation other than monte carlo variation and mismatch. So ignoring those, the equation in the model is just (l * rsheet / w) which would be (0.5 / 0.35) * 2000 = 2857 ohms. @User: Can you confirm what's in the netlist?