On FLEURS Hindi, the word error rate of GPT-4o mini Transcribe was about 57 percent lower than that of Whisper large-v3, in the Voxtral paper's comparison.

Checked against its source on October 17, 2026

What it measures: Word error rate on the FLEURS Hindi test set, from the per-language table in the Voxtral paper. The percentage is the relative gap between the two error rates.
How it is calculated: (28.87 - 12.36) / 28.87 = 0.57
What it does not show: One language on one benchmark, in a comparison run by the authors of the Voxtral paper. Other languages in the same table do not show the same pattern.

Calculated from