GMP is the most widely used for large numbers. The precision is arbitrary. The floating point class is precise up to 41 billion digits.

I'm pretty sure it's integer classes don't have a limit (other than your RAM of course.)