Generally, passive mixers would have higher linearity . However, higher noise figures and negative gain are major drawbacks. So, in general, it'd not be preferable to use a passive mixer as the 1st mixer in the Rx chain ( or as a direct down-conversion mixer ) or otherwise overall system noise figure would be degraded. It'd be more suitable for use where the signal is healthy.
Gilbert Cell mixer has lower linearity in general. However, it's possible to get reasonable gain from it. it requires higher supply voltages of course. Low voltage versions are usually pseudo-differential ( tail source is eliminated ).
However, it's very popular and well-stablished with tons of papers/references.