Whisper large-v3-turbo

Name: Whisper large-v3-turbo
Author: OpenAI

OpenAI's faster Whisper speech recognition model: large-v3 with the decoder reduced from 32 to 4 layers, much quicker with small accuracy loss.

Official page

Run it locally

Memory: About 2–3 GB of GPU memory, or runs on CPU with whisper.cpp
Formats: Safetensors (Hugging Face), whisper.cpp (GGML)

Overview

Whisper large-v3-turbo is a pruned version of large-v3 released in October 2024. It transcribes speech in many languages much faster at a small cost in accuracy.

Highlights

About 809M parameters
Multilingual transcription
Much faster than large-v3
MIT license

Typical uses

Local transcription of meetings, interviews and podcasts.

Things to consider

Not trained for translation, unlike other Whisper models.

Specifications are taken from the model card on Hugging Face and the provider’s announcement. Benchmark results depend on evaluation settings; see how AI benchmarks work.

#Speech Models #Open Weights #OpenAI