My experience: the k-3 had worse noise than the k-5 when pixel peeping, but for the overall image performed better because of the additional sharpness due to lack of AA filter. I'd be happy using the k-3 at ISO 3200, but was only ever happy with 2000 for the k-5. I had a ks-1 for a while, the ISO performance was a touch better than the k-3. I'd probably prefer to keep that at 3200, but 4000 was there if needed. The downside though is that the white balance and colour accuracy wasn't as good.
The k-1 was a bit of a jump. I personally prefer to keep it under ISO 10000, but if it's warrented, I'll push it to ISO 16000, but that will require a bit more effort in light room. At pixel peeping resolutions it's somewhere between half a stop to a full stop better than the k-5 at least. When however you consider that you're getting 36mp rather than 16mp, the overall effect is that the image quality at higher isos is improved (unless you're uploading the images to Facebook, which seems to amplify the noise in a really nasty way!)
Comparing ISO performance between the bodies, I'd say that ISO 1600 on the k-5, is equivalent to ISO 2000 (k-3), ISO 2500 (ks-1), ISO 8000 (k-1). [when viewing the final images at full size, and not pixel peeping]. I've not had a chance to use the KP myself, but if it slotted in somewhere between 2500 and 3200, it wouldn't suprise me.
But that's just my opinion. The difference between the apsc bodies isn't massive, the difference between the apsc bodies and the k-1 is noticeable.
(If the k-1 was in crop mode, it's not that different to the ks-1 in my opinion - that apsc camera being the best ISO performer I've played with to date - although the white balance and colour rendition isn't up there with any of the 14bit raw models)
Last edited by robthebloke; 04-30-2017 at 08:03 AM.