The Sony a7r iii has a well-earned reputation for low-light shooting. It has a superb full-sized sensor and a wide ISO range, going from 50 through 102400. But that doesn't mean that the top end of the ISO range has the same image quality as at the low end.
You would naturally expect a few areas to suffer. The first is noise and/or mottling. Post-processing noise reduction apps aren't quite the post-processing staple that they used to be, but that doesn't mean that sensor noise isn't an issue anymore. A second area that suffers is dynamic range. The Sony is neck and neck with the Nikon D850 in leading the pack in terms of dynamic range. The third aspect that suffers is color rendition. At the very top end, colors tend to get washed out, less accurate, and sometimes end up in strange casts.
As you go higher up the ISO scale, all of these areas that affect image quality tend to suffer, and that's something that's standard to all digital camera sensors. The question becomes, "just how bad is it?"
Sony a7r iii ISO Sample Images
DXOMark has some great technical performance tests on the Sony a7r iii's ISO. What I wanted to do is a different approach, looking at real-world, practical examples.
Here's a collection of images across the range of available ISO values on the a7r iii. These were all shot in aperture priority mode on a tripod with manual ISO settings. They were shot in RAW+JPG, and you can download both versions for each image in the links in the captions. Other settings were automatic white balance, Adobe RGB colorspace. There hasn't been any post-processing applied, and you download the original JPGs as they were straight out of the camera.
To give the best idea of what the camera does, I deliberately haven't done anything to improve the images or mitigate the effects of the high ISO settings. But that's obviously something you can tackle in software like Lightroom or any of the other image processing apps, especially if you shoot in RAW.
And with these, I deliberately shot something with smooth tonal gradations from light to dark so as to show up any issues with noise. It's also a good subject to test dynamic range.
With the small versions embedded below it's hard to see any real differences aside from the obvious color issues at the high end, so I've included links to the original straight-out-of-camera versions of both the JPG and RAW files.