For dictation on a Mac, use Whisperer V3, the default model. It's Whisper Large V3 Turbo, quantized down to about 547 MB, so it loads quickly, runs fast on the Apple Silicon GPU and keeps Large-class accuracy across 97 languages. Move to Large V3 only when you need maximum accuracy on hard audio, and drop to Small or Base only if memory is tight.
Whisperer offers several Whisper models in Settings. Here's how they differ.
The models#
Sizes are the download sizes shown in Whisperer.
| Model | Size | Speed | Best for |
|---|---|---|---|
| Whisperer V3 (Large V3 Turbo, Q5) | 547 MB | Fast | The default. Everyday dictation in any of 97 languages |
| Large V3 Turbo | 1.5 GB | Fast | The same model at full precision |
| Large V3 Q5 | 1.1 GB | Medium | Large V3 accuracy in less space |
| Large V3 | 2.9 GB | Slowest | Maximum accuracy, accents, noisy rooms |
| Distil Large V3 | 756 MB | Very fast | Fast English-heavy dictation |
| Medium | 1.5 GB | Slow | Older choice, mostly replaced by Turbo |
| Small | 466 MB | Medium | Low memory |
| Base | 142 MB | Fast | Very low memory, short notes |
| Tiny | 75 MB | Fastest | Testing only |
There's also a Hebrew Large V3 Turbo model (1.5 GB) for people who dictate mostly in Hebrew.
Recommendations#
Most people: Whisperer V3. It's what Whisperer downloads on first launch. It's quick enough to drive the live preview while you speak and accurate enough that you rarely fix anything.
Maximum accuracy: Large V3. Worth it for strong accents, technical vocabulary, noisy rooms, or languages where smaller models struggle. It's slower and takes more memory, so it suits Macs with 16 GB or more.
Tight on memory: Small or Base. On an 8 GB Mac with lots of apps open, a smaller model keeps things responsive. Expect more mistakes on long or fast speech.
Whisper isn't the only engine. Settings also has Parakeet (25 languages, on the Neural Engine), Qwen3 ASR (31 languages) and WhisperKit (Core ML), all offline. Apple Speech drives the live preview. See Whisper vs Parakeet vs Apple Speech.
Language matters#
Whisper handles English well at every size. Other languages benefit from the larger models:
- Spanish, French, German, Portuguese: Whisperer V3 is plenty.
- Chinese, Japanese, Korean, Arabic: Whisperer V3 or Large V3.
- Less common languages: Large V3 is the safest bet.
Whisperer detects the language automatically. If you only ever dictate in one language, choosing it in Settings can help with very short phrases.
Tips that matter more than model size#
- Use a decent microphone. Input quality beats model size. A headset mic close to your mouth helps more than moving to Large V3.
- Add names to the dictionary. The personal dictionary fixes the words any model gets wrong, like product names and colleagues' surnames.
- Close memory-hungry apps before running Large V3 on a 16 GB Mac.
- Keep the model warm. Whisperer keeps the model loaded so the first dictation after a pause starts quickly.
Switching models#
You can download several models and switch in Settings without restarting the app. Many people keep Whisperer V3 as the default and use Large V3 for file transcription, where speed matters less.
FAQ#
Which Whisper model should I use for dictation?#
Start with Whisperer V3, the default. It's Whisper Large V3 Turbo compressed to about 547 MB, fast on Apple Silicon and accurate across 97 languages. Switch to Large V3 only if you need the last bit of accuracy and have memory to spare.
Does a bigger Whisper model always mean better dictation?#
Not always. Large V3 is the most accurate on hard audio, but it's 2.9 GB and slower. For everyday dictation a good microphone and a personal dictionary make more difference than moving from Whisperer V3 to Large V3.
Related: Whisper vs Parakeet vs Apple Speech, Getting started. See pricing and all dictation features.