Benchmarks against distil-whisper/distil-large-v3?

#40
by datasaurus - opened

Does anyone have any latency metrics comparing v3-turbo against distil-whisper/distil-large-v3?

Not an exhaustive test, but on an RTX 3090 with flash attention 2, 100 minutes of audio:
distil-whisper/distil-large-v3 = 2 m 17s
openai/whisper-large-v3-turbo = 2m 59s

Sign up or log in to comment