whisperkit-coreml
OpenAI Whisper in CoreML format, published by Diction Labs for on-device speech-to-text on Apple hardware. Everything here is our own build, made with whisperkittools (MIT) from OpenAI's MIT-licensed Whisper checkpoints.
Use with WhisperKit
let config = WhisperKitConfig(
model: "openai_whisper-base",
modelRepo: "DictionLabs/whisperkit-coreml"
)
let pipe = try await WhisperKit(config)
Runs on argmax-oss-swift (MIT). Tokenizers are
resolved separately from the matching openai/whisper-* repo, so a first run still needs network
access even with the weights already on disk.
Variants
| Folder | Base model | Size |
|---|---|---|
openai_whisper-base | openai/whisper-base | full precision |
openai_whisper-small | openai/whisper-small | full precision |
openai_whisper-large-v3-turbo | openai/whisper-large-v3-turbo | full precision, 1.63 GB |
dictionlabs_whisper-large-v3-turbo-q6q8 | openai/whisper-large-v3-turbo | compressed, 737 MB |
dictionlabs_whisper-large-v3-turbo-q6q8 is a compressed version of the turbo model above,
same weights, reduced precision (encoder 6-bit, decoder 8-bit), 45% of the size. Verified against
the full precision build with real transcription tests, not just internal accuracy checks.
Headline (2026-09-04): on 150 clips of real spontaneous speech (VoxPopuli, 113 speakers,
en/de/fr) the compressed build scores 11.14% WER vs 10.96% for full precision — a +0.18 pt
difference inside sampling noise — and produces the identical output to full precision on
141 of 150 clips. Under added noise down to 0 dB SNR (240 clips) the two are within 0.1 pt.
Runs at ~9.6× realtime on an M1 Mac mini via the Neural Engine.
Full methodology, noise tables, weight-fidelity figures and caveats:
dictionlabs_whisper-large-v3-turbo-q6q8/README.md.
Read-speech check (FLEURS, 5 clips per language, absolute numbers inflated by a simple normalizer):
| Language | Full precision (WER/CER) | Compressed (WER/CER) |
|---|---|---|
| German | 2.8% | 3.4% |
| English | 15.5% | 15.5% |
| French | 7.8% | 8.8% |
| Japanese | 2.6% | 2.6% |
| Korean | 18.1% | 22.5% |
| Chinese | 9.0% | 9.0% |
Licence
MIT, same as upstream Whisper.