Well, because it uses a content-addressed hash, you don't need to manage .torrent files or anything.

Also, if people make two torrents with the same files, and you download one of them, it only downloads from people who seed that torrent, not from people seeding the other torrent with the same files. Since IPFS has content addressed hashes each file, you can download from anyone who has that file.

Its quite a bit more, but I wouldn't blame you for making the comparison.

Side note: unless you're a dev really trying to get in on bleeding edge stuff, ignore the hype. The IPFS team is doing great, but there's a lot of work left to do.

The torrent TLDR is; theres a lot of quality-of-life things missing from torrents (if they were going to be loaded like webpages) and it turns out fixing some of those quality of life things requires solving some pretty hard technical challenges, even if they don't feel much different to the end user. Things like decentelized discovery, content address hashing (best thing since sliced bread), merkel trees for de-duplication, and change detection.