Fact · Speech recognition
12.3percent word error rate
In the same 2025 9to5Mac test of one 7-minute-31-second episode, NVIDIA's Parakeet v2 had a word error rate of 12.3 percent, higher than Apple's 10.3 percent.
Checked against its source on October 16, 2026
In the source's own words
“Parakeet v2 2 seconds 5.8% 12.3% Whisper Large V3 Turbo 40 seconds 0.2% 1.5% Apple 9 seconds 1.9% 10.3%”
How accurate is Apple's new transcription AI? We tested it against Whisper and Parakeet, Marcus Mendes, 9to5Mac, 2025, Results table, 'How they did'.
Open the original- What it measures
- The author timed each engine and scored it on character error rate and word error rate, using two Hugging Face scoring tools. The table lists the time, then the character error rate and word error rate for each engine.
- What it does not show
- One author, one recording of about eight minutes, run on one Mac. It does not show how the engines compare on other audio, languages or microphones.
Source
Cited on
- iPhone Dictation Not Working? How to Fix ItGuide
- What Is Whisper AI? OpenAI's Speech Model ExplainedGuide
- Best Dictation Apps for iPhone and AndroidGuide
- Speech to Text Without AI: Dictation With No AI RewriteGuide
- Open-Source Speech Recognition: The Best ModelsGuide
- Captions and Subtitles: Formats, Rules and ToolsGuide
- How Apple Dictation Works: On-Device, Server and AIGuide
- Is AI dictation more accurate than built-in dictation?Question
- Is Apple Dictation any good?Question
- Why do speech-to-text accuracy claims disagree?Question
- Why is Apple Dictation so bad?Question