Whisper large-v3
OpenAI · Speech to text · 1.55B · Released 6 November 2023
Commercial use permitted
Open weights
Runs on CPU
Apple Silicon
OpenAI's flagship open speech-to-text model, and the de facto standard for local transcription. It handles many languages, is robust to accents and background noise, and runs well through whisper.cpp or faster-whisper. MIT licensed.
Strengths
- Strong, robust transcription across many languages
- Efficient community runtimes (whisper.cpp, faster-whisper) run it on modest hardware
- MIT licensed, with no commercial-use conditions
Weaknesses
- Can hallucinate text during long silences or noisy passages
- Processes audio in 30-second windows, so very long files need chunking
- Real-time use needs a smaller or distilled variant on modest hardware
Hardware requirements
| Quantisation | Approx. VRAM | Notes |
|---|---|---|
| Int8 | ~1.5GB | Common for faster-whisper, minimal accuracy loss |
| FP16 | ~3GB | Full precision |
Also runs on CPU (slower). Optimised builds available for Apple Silicon.
Licence
MIT — read the licence
Availability
Recommended for
- Local transcription and subtitling
- Multilingual speech-to-text
- Feeding audio into a local RAG or notes pipeline
Catalogue entry last verified 30 July 2026. Specifications change; verify anything you are about to spend money on.