Fact · Speech recognition
2.18percent word error rate
NVIDIA Canary 1B v2 had a word error rate of 2.18 percent on the LibriSpeech test-clean test set in the Open ASR Leaderboard table.
Quote found word for word on the live source, last checked October 19, 2026
In the source's own words
“| **WER ↓** | **RTFx** | **Mean** | **AMI** | **GigaSpeech** | **LS Clean** | **LS Other** | **Earnings22** | **SPGISpech** | **Tedlium** | **Voxpopuli** | ... | `Canary-1b-v2` | 749 | 7.15 | 16.01 | 10.82 | 2.18 | 3.56 | 11.79 | 2.28 | 4.29 | 6.25 |”
canary-1b-v2 model card (raw README), NVIDIA, Hugging Face (NVIDIA model card), 2025, Open ASR Leaderboard table, LibriSpeech test-clean column.
Open the original- What it measures
- The model card's Hugging Face Open ASR Leaderboard table gives the WER for each dataset in the leaderboard's English set.
- What it does not show
- A vendor-run score on the leaderboard's English datasets, with the leaderboard's text normalization. Compare only with systems scored the same way.
In the data
A row of Speech Recognition Accuracy: Word Error Rates, as Canary 1B v2, LS test-clean English: 2.18 percent.
- System
- NVIDIA Canary 1B v2
- Kind
- Open model
- Test set
- LibriSpeech test-clean
- Language
- English
- Speech
- Clean read speech
- Year
- 2025