About the output compressed files

[lyysl]

I am new to RGB compression. I have the question of whether RGB compression should output a compressed image that has a smaller output size or a compressed file from the encoder, which I can later pass to the decoder to get the reconstructed image?

What is the point of getting a lossy output with similar file size?
Would it be possible to save e.g. the latent representation from the encoder?

Thank you.