Fact · Speech recognition
4.1times faster
On the faster-whisper project's CPU benchmark with the small model, faster-whisper with int8 precision took about 4.1 times less time than openai/whisper in fp32 precision.
Checked against its source on October 17, 2026
- What it measures
- Both runs transcribed the same 13-minute recording on the same CPU, according to the project's benchmark table.
- How it is calculated
- 418 seconds / 102 seconds = 4.1
- What it does not show
- A single benchmark run by the project that makes faster-whisper, using int8 precision for it and fp32 for openai/whisper. The project's own headline figure of up to 4 times is the more cautious number to quote.