Fact · Transcription
46.75percent word error rate
On the Mozilla Common Voice Hindi test in the Voxtral paper, Whisper large-v3 had a word error rate of 46.75 percent.
Checked against its source on October 17, 2026
In the source's own words
“Model ar nl en fr de hi it pt es Whisper large-v3 50.58 5.83 22.91 11.33 6.25 46.75 6.81 7.17 5.66”
Voxtral, Mistral AI, arXiv (Audio and Speech Processing), 2025, Table 5 (Mozilla Common Voice), column hi.
Open the original- What it measures
- Common Voice is speech recorded by volunteers on their own devices, so its audio is less controlled than the read speech in FLEURS. The table is in the same paper as the FLEURS table, so the two can be compared.
- What it does not show
- A single paper's measurement on one test set. It shows that the same model can score very differently on two Hindi test sets.