@User: It works as long as you don't use the "nf" parameter. Ngspice has a bug in which it does not select bins properly based on the number of fingers. If you use "m" instead of "nf" then it should be okay.

Am I right in thinking that m represents multiple devices whereas nf would represent a single device with multiple fingers? I would expect multiple 1-finger devices to have more source and drain diffusion capacitance than one multiple finger device (where there is a gate on each side of the fingers of drain diffusions so there is half as much drain diffusion per gate width) For modelling a layout with good high-speed or RF performance, do we need nf working? or, are the s,d area and perimeter passed separately so that it wouldn't matter?

@Chris Jones: Yes, although "m" is a general-purpose parameter in SPICE and can apply to any device or subcircuit, while "nf" is a parameter specific to bsim MOSFET models (possibly only bsim4?). "nf" has the somewhat oddball definition that when it is used, "w" represents the total device width (sum of widths of all fingers), whereas if you use "m", then "w" represents the width of a single device. There is a patch to ngspice branch pre-master that was made two weeks ago, but I have not checked back in with Holger Vogt to see if there has been any update, or if has been merged into the master branch.