Fact · Speech recognition
8.04percent word error rate
In the Voxtral paper's Table 2, Gemini 2.5 Flash had a macro-average word error rate of 8.04 percent across the English short-form test sets.
Quote found word for word on the live source, last checked October 19, 2026
In the source's own words
“Gemini 2.5 Flash 88.64 ± 2.28 4.54 ± 0.07 8.04”
Voxtral, Mistral AI, arXiv (Audio and Speech Processing), 2025, Table 2, En Short WER column.
Open the original- What it measures
- Table 2 compares Voxtral checkpoints after supervised finetuning and after preference training. The English short-form figure is a macro-average across its test sets.
- What it does not show
- One paper's runs under its own text normalization and decoding, so compare systems within this table, not with other papers' numbers. The paper gives no date per row.
In the data
A row of Speech Recognition Accuracy: Word Error Rates, as Gemini 2.5 Flash, EN short-form avg: 8.04 percent.
- System
- Gemini 2.5 Flash
- Kind
- Cloud service
- Test set
- English short-form average (Voxtral paper)
- Language
- English
- Speech
- Mixed benchmark
- Year
- 2025