There's (at least) two things going on here:
1. One is 8bit, the other 16bit grayscale
2. The smaller file size one is much smoother, meaning the lossless compression algorithm is much more effective. Conversely, there's a lot of finer, random, detail in the larger map.
Here's a link to an 8bit version of Tricorn - it's still really big in comparison (30mb): https://mega.nz/file/6qAFCKwY#HODbEEbjJgZAD43Yuq4FTrzTycI27oF-flvCWYLDUOo