Fact · Dictation software
4decoder layers
The Whisper maintainers say large-v3-turbo is an optimized version of large-v3 with only 4 decoder layers, down from 32 in the large series.
Quote found word for word on the live source, last checked October 19, 2026
In the source's own words
“It is an optimized version of Whisper large-v3 and has only 4 decoder layers”
turbo model release (openai/whisper discussion #2363), OpenAI whisper maintainers, GitHub, openai/whisper repository, 2024, Release post.
Open the original- What it measures
- A description of the model architecture from the maintainers' release post.
- What it does not show
- Architecture is a design detail. It does not by itself show speed or accuracy.