Fact · Voice and AI
24percent lower word error rate
On the Open ASR Leaderboard's English short-form tests, the average word error rate of Canary-Qwen 2.5B was about 24 percent lower than that of Whisper Large v3.
Checked against its source on October 17, 2026
- What it measures
- Average word error rate over the English short-form tests of the Open ASR Leaderboard. The percentage is the relative gap between the two averages, not a difference in points.
- How it is calculated
- (7.44 - 5.63) / 7.44 = 0.24
- What it does not show
- A ratio of two averages on one leaderboard. The leaderboard's normalized scoring removes punctuation and fillers, so this gap does not measure formatting or cleanup.
Calculated from
- On the same leaderboard track, OpenAI Whisper Large v3 had an average word error rate of 7.44 percent and an inverse real-time factor of 146.
- On the same leaderboard track, NVIDIA Canary Qwen 2.5B, a speech encoder with a language-model decoder, had an average word error rate of 5.63 percent and an RTFx of 418.