No. Bill's seems to handle unicode better. Johan Björklund's glue cmod (to json-c, http://oss.metaparadigm.com/json-c/) probably outperforms and outstyles us both in speed and quality (assuming the lib is good), and even throws mildly informative errors on illegal input. Unicode (wide strings) unfortunately belong to that category too, though. :-/
From comments in Bill's code it seems to be based on some C#
implementation by Are Bjolseth, in turn based on something else, but I didn't manage to look at it for very long periods of time; its inlined refdocs makes one want to gnaw off a leg. :-)
I think that is about the evaluation I can muster at the moment. The three obviously aim a bit differently. For myself I wanted something that could re-layout things. Json-c ought to be good at churning large volumes of latin-1 (if there is any merit with it being written in C, at least), but does not easily adjust to change formatting, in case its default output looks (whatever those may be) are not what one wanted.
I think I'll try to pipe Bill's code through cpp or something and perhaps cherry pick the string formatter, if I keep working on mine.