Fact · Speech recognition
9.87percent word error rate
NVIDIA Canary 1B v2 had a word error rate of 9.87 percent on the This American Life long-form recordings.
Quote found word for word on the live source, last checked October 19, 2026
In the source's own words
“| **Dataset** | **WER ↓** | ... | This American Life | 9.87% |”
canary-1b-v2 model card (raw README), NVIDIA, Hugging Face (NVIDIA model card), 2025, Long-form inference table, This American Life row.
Open the original- What it measures
- The card's long-form inference section gives one WER per recording set for whole recordings.
- What it does not show
- A vendor-run figure on two recordings, not a broad survey of long-form speech.
In the data
A row of Speech Recognition Accuracy: Word Error Rates, as Canary 1B v2, This American Life: 9.87 percent.
- System
- NVIDIA Canary 1B v2
- Kind
- Open model
- Test set
- This American Life
- Language
- English
- Speech
- Mixed benchmark
- Year
- 2025