The first real third-party benchmark of Apple’s SpeechAnalyzer API just hit 369 points on HackerNews, and the result is blunt: for English transcription on current iPhones and Macs, “just use Whisper” is no longer the default answer.
The numbers
A 34-minute 4K video: SpeechAnalyzer transcribed it in 45 seconds. Whisper Large-v3 Turbo took 101 seconds. That’s 55% faster — on the same machine. Per second of audio, Apple’s engine runs about 3x faster than Whisper Small while beating it on accuracy: word error rate drops from 9.02% to 2.12% on clean English speech, 16.25% to 4.56% on noisy audio.
What developers actually get
SpeechAnalyzer/SpeechTranscriber is Apple’s new system-level speech-to-text API, replacing the old SFSpeechRecognizer. It ships free with iOS 26/macOS 26 — on-device, no model downloads, no per-minute fees. A few lines of Swift gets you meeting notes, voice memo transcription, or video captions.
The catches are real: roughly 30 language locales versus Whisper’s ~100, Apple platforms only, and proper nouns plus heavy accents still trip it up.
The bigger signal: every transcription app whose pitch was “fast local Whisper” just had a moat layer pulled out by the OS itself.
You Might Also Like
- Oculos Turns Every Desktop app Into a Json api and its Exactly What ai Agents Needed
- Ollama mlx on Apple Silicon 1810 Tokens sec Prefill and the end of Llama cpp on mac
- Tinygpu George Hotz got Apple to Sign an Nvidia gpu Driver for mac
- Openai gpt Realtime 2 Translate Whisper Three Voice Models one api Several Startups Erased
- Apple afm 3 Runs a 20b Model on Your Iphone and Hands it to Developers
Top AI Product 
Leave a comment