GPU Instancing Batch Limits?

Hi, I am instancing some cubes and it seems the batching ends at 3000 vertices / 4500 indices, for a total of 125 cubes per batch. Is this typical, or is this platform specific? I am getting these results on OSX and Android. I was thinking that the limts would be higher.

Nevermind! #define UNITY_MAX_INSTANCE_COUNT 1000