When Tidy processes short text (100–500 tokens), inference latency is approximately 1–2 seconds for initial loading plus 200–800 milliseconds of computation, with continuous use impacting battery life by less than 5% per hour.