The major difference is in the Garbage Collection techniques.

Python uses a not so reliable and a much slower approach called reference counting. The reason why it’s slow is that each time an object is referenced and dereferenced, the interpreter/VM must check to see if the count has gone down to 0. This is time-consuming. It’s also unreliable because of its inability to address the circular reference issues (google it). One thing to note is that the dead objects are immediately removed unlike in Java. Also, it uses much less memory to perform this action.

Java, on the other hand, employs a much more logical approach called CMS (concurrent mark sweep) which basically checks for all ‘reachable’ code. Anything that can’t be reached is marked as dead code. JVM then moves these dead objects in fresh memory and, at a later point, sweeps the entire memory region in one sweep. It could also run in a separate thread unlike in python. There are variations to above approach and these garbage collectors are much more optimized. It is faster than python but uses more memory to perform all the above actions.