Fact · Speech recognition
7.381428571428572percent average WER
On the Open ASR Leaderboard's English short-form results, soundsgoodai/Zipformer-cr-ctc-transducer-XL-290M (fast-gpu-asr, transducer_modified_beam_search) had an average word error rate of 7.38 percent.
Quote found word for word on the live source, last checked October 19, 2026
In the source's own words
“transducer_modified_beam_search)",7.381428571428572,19289.67”
Open ASR Leaderboard: English short-form results (english_short_latest.csv), hf-audio (Hugging Face), Hugging Face, hf-audio organization, 2026, english_short_latest.csv, column avg.
Open the original- What it measures
- The average is over the leaderboard's English short-form test sets, under the leaderboard's shared text normalization. The RTFx column is inverse real-time factor, a speed measure.
- What it does not show
- An average over several English test sets, not one test set. Results depend on the normalization and the test sets, and change as the leaderboard is updated.
In the data
A row of Speech Recognition Accuracy: Word Error Rates, as Zipformer XL, transducer beam search: 7.38 percent.
- System
- soundsgoodai/Zipformer-cr-ctc-transducer-XL-290M (fast-gpu-asr, transducer_modified_beam_search)
- Kind
- Open model
- Test set
- Open ASR Leaderboard (English average)
- Language
- English
- Speech
- Mixed benchmark
- Year
- 2026
- Note
- Kind follows the leaderboard's License column (license: cc-by-nc-4.0): Proprietary is Cloud service, any other license is Open model.