This question is a FAQ.
differentiate, integrate, factorise, complete the square, give results in exact form, all the basics, etc.
Actually, if that's all you need, the open source and freely available package maxima will do fine. Maple and maxima "have a common ancestor", by the way, so their syntax is similar.
mgb_phys said:
Traditionally Mathematica has been used for more pure maths sort of tasks and Maple has been aimed at modelling and analysing data.
For some Groebner basis type tasks, many practioners might give the edge to Maple. I would have said that MATLAB is more likely to be used for many common modeling tasks, especially for large scale linear algebra problems.
One important feature of Maple is that while this is not free-ware, the source code is freely available, which is of paramount importance to careful researchers. I know quite a few people who have been bitten by mysterious Mathematica bugs, although to be fair, all complicated software packages have bugs. For this reason, careful researchers will try to maintain proficiency in at least two general purpose symbolic computation packages, and to check results one against the other. With some awkwardness it is possible to port data between Maple and Mathematica; e.g. Maple has a tool which converts Mathematica routines to Maple routines.
OTH, everyone who has used both will probably agree that Mathematica has more attractive plotting (e.g. when preparing figures for a published paper).
mgb_phys said:
Both are very good but with their own quirks.
Agreed.
It might be worth mentioning that there are many excellent packages available for specialized computations, e.g. group theory or algebraic geometry. Some of these run under Mathematica or Maple; others are powerful symbolic computation systems in their own right (e.g. GAP, Macaulay2, Singular).