Fact · Speech recognition
10.1percent AA-WER
On Artificial Analysis's speech-to-text leaderboard, Whisper Large v3 Replicate large-v3 had an AA-WER of 10.1 percent, its word error rate on the index's three test datasets.
Quote found word for word on the live source, last checked October 19, 2026
In the source's own words
“Whisper Large v3 Replicate large-v3 10.1% 2.7 4.23 Details”
Speech to Text (ASR) Providers Leaderboard & Comparison, Artificial Analysis, Artificial Analysis, 2026, Summary of Key Metrics table.
Open the original- What it measures
- Non-streaming AA-WER v2 index: AA-AgentTalk (50 percent), VoxPopuli-Cleaned-AA (25 percent) and Earnings22-Cleaned-AA (25 percent), about 8 hours of audio. Each row is the provider's hosted endpoint for the model.
- What it does not show
- AA-WER is an index, not a single public test set, and one of its datasets is proprietary. The page does not name a language for the index; the v2.0 article builds its text normalizer on Whisper's English normalizer. Figures change as the leaderboard updates.
In the data
A row of Speech Recognition Accuracy: Word Error Rates, as Whisper Large v3 Replicate large-v3: 10.1 percent.
- System
- Whisper Large v3 Replicate large-v3
- Kind
- Cloud service
- Test set
- AA-WER
- Language
- English
- Speech
- Mixed benchmark
- Year
- 2026
- Note
- Language is not named on the leaderboard page; the AA-WER v2.0 article describes a normalizer built on Whisper's English normalizer.