Canary-1B-v2 & Parakeet-TDT-0.6B-v3: Efficient and High-Performance Models for Multilingual ASR and AST

Open paper
Benchmark Model Rank Results
speech-recognition-on-aminvidia/parakeet-tdt-0.6b-v3#16Word Error Rate (WER): 10.58
speech-recognition-on-aminvidia/canary-1b-v2#27Word Error Rate (WER): 14.71
speech-recognition-on-ami-cleanednvidia/parakeet-tdt-0.6b-v3#16Word Error Rate (WER): 9.41
speech-recognition-on-ami-cleanednvidia/canary-1b-v2#30Word Error Rate (WER): 13.02
speech-recognition-on-earnings-22nvidia/parakeet-tdt-0.6b-v3#19Word Error Rate (WER): 10.77
speech-recognition-on-earnings-22nvidia/canary-1b-v2#25Word Error Rate (WER): 11.35
speech-recognition-on-gigaspeech-cleanednvidia/parakeet-tdt-0.6b-v3#11Word Error Rate (WER): 8
speech-recognition-on-gigaspeech-cleanednvidia/canary-1b-v2#41Word Error Rate (WER): 9.12
speech-recognition-on-gigaspeech-testnvidia/parakeet-tdt-0.6b-v3#10Word Error Rate (WER): 8.07
speech-recognition-on-gigaspeech-testnvidia/canary-1b-v2#40Word Error Rate (WER): 9.2
speech-recognition-on-librispeech-test-cleannvidia/parakeet-tdt-0.6b-v3#34Word Error Rate (WER): 1.51
speech-recognition-on-librispeech-test-cleannvidia/canary-1b-v2#53Word Error Rate (WER): 1.76
speech-recognition-on-librispeech-test-othernvidia/canary-1b-v2#26Word Error Rate (WER): 3.07
speech-recognition-on-librispeech-test-othernvidia/parakeet-tdt-0.6b-v3#28Word Error Rate (WER): 3.12
speech-recognition-on-spgispeechnvidia/canary-1b-v2#10Word Error Rate (WER): 2.03
speech-recognition-on-spgispeechnvidia/parakeet-tdt-0.6b-v3#47Word Error Rate (WER): 3.63
speech-recognition-on-voxpopulinvidia/parakeet-tdt-0.6b-v3#15Word Error Rate (WER): 5.89
speech-recognition-on-voxpopulinvidia/canary-1b-v2#16Word Error Rate (WER): 5.93
speech-recognition-on-voxpopuli-cleanednvidia/parakeet-tdt-0.6b-v3#15Word Error Rate (WER): 3.19
speech-recognition-on-voxpopuli-cleanednvidia/canary-1b-v2#45Word Error Rate (WER): 4.39

Rank counts only results with a code link.