Is it better to use "is" or "==" for number comparison in Python?

Use ==.

Sometimes, on some python implementations, by coincidence, integers from -5 to 256 will work with is (in CPython implementations for instance). But don't rely on this or use it in real programs.

That will only work for small numbers and I'm guessing it's also implementation-dependent. Python uses the same object instance for small numbers (iirc <256), but this changes for bigger numbers.

>>> a = 2104214124
>>> b = 2104214124
>>> a == b
True
>>> a is b
False

So you should always use == to compare numbers.

Others have answered your question, but I'll go into a little bit more detail:

Python's is compares identity - it asks the question "is this one thing actually the same object as this other thing" (similar to == in Java). So, there are some times when using is makes sense - the most common one being checking for None. Eg, foo is None. But, in general, it isn't what you want.

==, on the other hand, asks the question "is this one thing logically equivalent to this other thing". For example:

>>> [1, 2, 3] == [1, 2, 3]
True
>>> [1, 2, 3] is [1, 2, 3]
False

And this is true because classes can define the method they use to test for equality:

>>> class AlwaysEqual(object):
...     def __eq__(self, other):
...         return True
...
>>> always_equal = AlwaysEqual()
>>> always_equal == 42
True
>>> always_equal == None
True

But they cannot define the method used for testing identity (ie, they can't override is).

>>> a = 255556
>>> a == 255556
True
>>> a is 255556
False

I think that should answer it ;-)

The reason is that some often-used objects, such as the booleans True and False, all 1-letter strings and short numbers are allocated once by the interpreter, and each variable containing that object refers to it. Other numbers and larger strings are allocated on demand. The 255556 for instance is allocated three times, every time a different object is created. And therefore, according to is, they are not the same.