AudioOpen weights
Whisper
OpenAI
Updated May 21, 2026 · Benchmarks via Artificial Analysis, specs & pricing via OpenRouter · Methodology · Report an error
Specifications
- Type
- Audio
- Access
- Open weights
- Released
- September 21, 2022
- License
- MIT
- Parameters
- 1.5B (large-v3)
- Input
- Audio
- Output
- Text
Our take
An open, robust speech-recognition and translation model trained on 680K hours of audio. Released under MIT, it became the default for transcription across countless products and research projects.