Whisper large-v3
OpenAIAUDIO Open weight
- Provider
- OpenAI
- Modality
- AUDIO
- Parameters
- 1.55B
- Context window
- —
- Weights
- Open
- Released
- 1 Oct 2024
Whisper is the free open-source STT standard — multilingual and self-hostable, though Indian-language specialists can beat it on Hindi.
Best for
- Free, self-hostable, high-accuracy multilingual transcription (99 languages)
- Privacy / offline transcription with full data control
- Batch transcription and subtitling at zero per-minute cost
- Robust to accents and noisy audio
How it compares — and the India angle
- Free and self-hostable — ideal for Indian developers; decent (not best) accuracy on code-mixed Hindi/Hinglish.
- For pure Indian-language STT, Sarvam Saaras often beats Whisper on Hindi/regional accuracy.
- Best free general-purpose STT; pair with Sarvam when Indian-language precision matters.
How to access
Open-weight, free to self-host (large-v3 ~1.55B). Also via OpenAI API (~$0.006/min).
Access Whisper large-v3Frequently asked questions
What is Whisper large-v3?
Whisper large-v3 is a audio AI model developed by OpenAI with 1.55 billion parameters.
Who made Whisper large-v3?
Whisper large-v3 was created by OpenAI.
Is Whisper large-v3 open source?
Yes, Whisper large-v3 has open weights and can be downloaded and self-hosted.
What is Whisper large-v3's context window?
The context window size for Whisper large-v3 has not been publicly disclosed.