Just because I was curious, the 0.01fA (10^-17A) works out to be all of 62.5 electrons per second. That's pretty impressive. Puling up the spec, though, reveals an accuracy of only 0.5% + 3fA and requires 1.4s to settle. So much of that 0.01fA is derived from resolution and noise analysis / reduction.
Anyways, op amps with < pA bias currents are not uncommon. For instance, consider the LMC6082, which is 10fA at room temperature, and about 2fA at 0C. Certainly you would want better than 1fA resolution to accurately characterize something <10fA.