Your basic assumption is false. Image filesize does not at all prove how high quality (=sharpness) it has. Blur an image and add grainy noise to it. Then save it with the same JPG compression level than the original. The result can have the same or more likely bigger filesize than the original.
Another example: An image with few details, for example one flat color surface can be compressed to a fraction of the filesize of an uncompressed image with no visible errors due the compression.
Example 3: A camera can produce uncompressed photos which have filesize about 20MBytes. If one makes them JPGs with compression level that causes less compression artefacts than the lens of the camera made errors, the filesizes of the JPGs can well be about 2MBytes depending on how complex the photographed subject was, but the compression did not reduce at all the subjective quality.
The only good way to compare the quality of 2 images without seeing them is to compare their pixel dimensions. A 100x100 px image have room for 10000 pixels of information and 1000 x 1000 px image has room for 1000000 pixels of information. The latter image has a potential to be much sharper.
But is it sharper depends on the content. If both have only the same flat single color, there's no difference. Also if the latter is made by resampling the smaller one to 1000% pixel dimensions, it can maximally be as sharp as the smaller, because there's no more information. The smoothing in the resampling (if used) more likely makes it subjectively very blurry altough it doesn't remove much information.
If the 1000x1000px image is the original, the content is a rich landscape and the 100x100 image is a resampled version, we can reliably say the 100x100px version is unsharper.
Thus comparing the pixel dimensions gives only some limit how high quality the image can possibly have if the image is complex, sharp and nobody has spoiled it. The ability to make estimates grows with the experience.
