The noise specifications quoted above shouldn't be compared to one another because they're derived very differently. When all you have is an A-weighted RMS equivalent noise measurement, you can generally estimate the audible noise level by adding 9 - 12 dB or so. That is, of course, assuming that the noise figure is a real measurement and not just a statement of a general goal--which is really how some manufacturers interpret the role of specifications, especially in the Far East.
When CCIR-weighted quasi-peak equivalent noise specs are available, they mean a heck of a lot more--but those are usually provided only for professional-quality microphones. Marketing folks adore "A"-weighted RMS measurements, which always look as if they're better (e.g. for the Neumann KM 183 it makes the difference between a noise spec of 13 dB versus 24). And a noise specification "at 1 kHz" doesn't mean anything at all; the terms simply don't go together that way.
--I would caution against the Røde NT4 not for reasons of noise but because it has coincident cardioid capsules at an included angle of only 90°, which isn't a generally useful setup for semi-distant stereo music recording. It's fine for stereo spot-miking, or for recording round-table meetings in an office--but for most people's purposes here, the pickup angle is just way, way too wide, and the results would be too much like mono. If they'd angled the capsules farther apart and/or made them not be coincident and/or given them a more sharply directional pattern, that might have been a lot better.
--best regards