None of them.
I know all the hard core mathematicians are going to recommend Spivak and they're all howling right now.But I can't in all good conscience recommend Spivak to a beginner in calculus, no matter how talented they are.
The problem with Spivak-and it is very beautiful,no question-is that it's just too austere and theoretical. The result is that even if a student is talented enough to master it, it'll only give half the story of calculus. Calculus is as much about the applications to the physical and social sciences as it is about the rigorous theory of limits and convergence of real valued functions. So a bizarre thing happens when a student learns calculus from this book-they can do epsilon-delta proofs with full rigor, but if you ask them to solve a simple projectile motion problem or to minimize an area, they look at you like you're speaking Martian. Choosing between a rigorous presentation of the theory of calculus and the study of it's applications to real life situations is no choice at all.
And the truth is-you shouldn't have to chose between them,there are now books that try and strive for a balance. Not many, but the ones that do exist are excellent.
An outstanding recent addition, which I'm dying to get my own copy of when I can, is Calculus With Applications by Peter Lax and Maria Shea Terell. It's a complete update and expansion of the calculus textbook Lax wrote in the 1970's and it's entire philosophy is to teach a course in single variable calculus that balances theory and applications in full measure. Not only does it prove all the major results of calculus fully and carefully, it contains an enormous number of applications and examples in mechanics, biology, chemistry, finance and the social sciences-as well as many major computer projects using computer algebra systems. If I could chose any text for an honors calculus class and could only pick one, this would be the one I'd pick.
Much more old fashioned but very much in this spirit is An Introduction to Calculus And Analysis by Richard Courant and Fritz John. In many ways, the Lax/Terrell book is an update of this one. Which is not really surprising since Lax first learned calculus in Germany from the original lectures of Courant at The University of Göttingen! It lacks computer exercises and doesn't quite have as many examples and applications as the Lax/Terrell book, but it's very similar otherwise and written by 2 masters. Still well worth reading after all this time.
Those would be my recommendations. If you insist on using Spivak, then I'd supplement it with the online version of Gilbert Strang's Calculus -which is the best applied calculus textbook that's ever been written.