Apple's new SpeechAnalyzer API, benchmarked against Whisper and its predecessor
Apple's new SpeechAnalyzer API, introduced with iOS 26 and macOS 26, has been benchmarked against Whisper models and the legacy SFSpeechRecognizer. On LibriSpeech test-clean, SpeechAnalyzer achieved a word error rate (WER) of 2.12%, compared to Whisper Small's 3.74%, Whisper Base's 5.42%, Whisper Tiny's 7.88%, and SFSpeechRecognizer's 9.02%. On the noisier test-other set, SpeechAnalyzer scored 4.56% WER versus Whisper Small's 7.95%, Whisper Base's 12.51%, Whisper Tiny's 17.04%, and SFSpeechRecognizer's 16.25%. The benchmark was run on an Apple M2 Pro with 32GB RAM running macOS 26.5.1, using identical production code paths. SpeechAnalyzer also ran roughly three times faster than Whisper Small. The source explicitly recommends migrating off SFSpeechRecognizer, as the new API cuts word error rate by 3.5 to 4x on the same audio with no accuracy trade-off.
Apple's new SpeechAnalyzer API offers a clear accuracy and speed upgrade over both Whisper and the legacy SFSpeechRecognizer.