From the model card
Fields below are copied from the tags and counters on the HuggingFace repository argmaxinc/whisperkit-coreml at our last fetch. They are set by the uploader, not verified by us; rows with no tag are omitted. How this page is made.
- Publisher (HF namespace)
- argmaxinc
- Pipeline tag
- automatic-speech-recognition
- Weight formats
- Core ML
- License tag
mit— read the license file in the repo before relying on it- Downloads (HF counter at last fetch)
- 11,227,615
- Likes (HF counter at last fetch)
- 205
- Model card
- https://huggingface.co/argmaxinc/whisperkit-coreml
Use cases
- On-device transcription for iOS/macOS apps without server-side ASR
- Privacy-preserving voice note transcription on Apple hardware
- Real-time caption generation in macOS applications
- Offline speech recognition in regions with unreliable connectivity
- Integrating ASR into Swift/Objective-C apps via the WhisperKit framework
Pros
- Runs entirely on-device — no network dependency or API cost
- Leverages Apple Silicon Neural Engine for efficient inference
- Multiple model sizes available (tiny through large) for different device capability levels
- Privacy-preserving by design — audio never leaves the device
Cons
- Apple platform only — no cross-platform use
- Requires WhisperKit framework integration in the host application
- Accuracy constrained by CoreML quantization vs. server-side full-precision Whisper
- Older or non-Apple Silicon devices see reduced performance
- Model downloads are bundled with the app or downloaded at first use — adds app size
Tags
whisperkitcoremlwhisperasrquantizedautomatic-speech-recognitionlicense:mitregion:us