Threaded connection model (no process spawning)
Undo-based MVCC (no need for vacuum)
InnoDB's use of a clustered index for PK (has pros/cons, but better for some workloads)
Ability to use alternative storage engines such as MyRocks (LSM based instead of B-tree; best-in-class compression)
Support for index hints (so query plans won't randomly change and bring your site down)
More mature logical replication (fully supports DDL, has no concept of limited "replication slots", etc)
That all said, there are also many areas where Postgres is better! Like all things in computer science, there are architectural trade-offs, and no single silver bullet is the best choice for all workloads.