Kind of . The minimum latency for the proper audio channels (i.e.: audio codec on Bela and Ctag), expressed in samples, is the built-in latency of the sigma-delta converter (about 40 samples, exact number depends on the converter), + ~2 samples in the McASP FIFOs, + double buffering (4 samples), which adds up to about 1 ms in total.
The audio expander cape can get to much smaller latency because the converters are not sigma-delta (but are - in turn - more noisy and don't reject alias), so they don't have the built-in delay. See here).