The importance of low and consistent latency in interactive music systems is well-established. So how do commonly-used tools for creating digital musical instruments and other tangible interfaces perform in terms of latency from user action to sound output? This paper examines several common configurations where a microcontroller (e.g. Arduino) or wireless device communicates with computer-based sound generator (e.g. Max/MSP, Pd). We find that, perhaps surprisingly, almost none of the tested configurations meet generally-accepted guidelines for latency and jitter. To address this limitation, the paper presents a new embedded platform, Bela, which is capable of complex audio and sensor processing at submillisecond latency.