Apple's new SpeechAnalyzer API, benchmarked against Whisper and its predecessor

Apple's new SpeechAnalyzer API significantly outperforms both its predecessor (SFSpeechRecognizer) and OpenAI's Whisper models in on-device speech recognition accuracy and speed, reducing word error rates by 3.5-4x compared to the legacy API while running three times faster than Whisper Small. For English transcription on current Apple devices, SpeechAnalyzer is now the most accurate on-device option available, though Whisper retains advantages in language support and cross-platform compatibility.