That's because you used dict. Inserting an element to a set or a dict is worst-case $$$O(n)$$$, and it's possible to construct a test case that causes this (read more about anti-hash-table tests).

Implementation of set and dict is different in Python and PyPy, and it's possible that an anti-hash test case only works in PyPy (or vice versa).

FYI, here's the hack test case used here.

95% of my submissions are in PyPy not Python normally coz it is nearly always faster and I got this problem ACC yesterday with your same idea nearly, the thing I am doing in these problems that hates Python are always look at a[i] if it is a[i] <= n or so, don't use dictionary just use array, when it is not maybe its a different solution or if I really wanna do hashing I will do d[str(a[i])] = "bla bla", coz I made a similar blog before and someone told me that and it worked, so use it always but be careful about this and switch to cpp when it comes to recursion and graph things or too tight problems and big arrays.

In this particular exercise you don't need to use dict. You can keep track of indexes with arrays 335981674

Sir, I have always been told that PyPy is faster than Python 3. Your case is truly a mystery.

I love you pypy.