Running 18 18 WhisperKit Benchmarks π Evaluate speech recognition model performance across devices and datasets