AFAIK, there are no standard benchmarks for API methods, and in fact there may be various JVM-based implementations that you use. Couple that with JVM JIT optimizations, garbage collection and many other things, and I doubt you can get globally significant numbers. Most of what you can do is write your own tests.
Some methods determine the computational complexity of operations in their JavaDocs. Some other methods describe other performance issues. Make sure you know about them and listen to them.
But besides this, most likely you are doing premature optimizations. Use the profiler to see that this is actually a bottleneck.
, , .. , , . , - , .