Question · Languages and accents
How accurate is speech recognition in Finnish?
On our Mac benchmark, Finnish read speech is transcribed with a word error rate of 7.73 percent by Whisper large-v3-turbo and 10.11 percent by Parakeet. That ranks Finnish eleventh of the 25 languages tested.
Updated October 19, 20263 min read11 cited facts, each traced to its source
Finnish is one of the better-served smaller languages: with the right engine only about 1 word in 13 is wrong on read speech. In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo had a word error rate of 7.73 percent 1 and Parakeet TDT 0.6B v3 had 10.11 percent 2, on 100 Finnish recordings from the FLEURS test set. We have no Apple SpeechTranscriber result for Finnish.
Whisper beats Parakeet in Finnish, the reverse of its neighbors
Finnish is a language where Whisper is the more accurate engine, by a clear margin of a couple of points. In Estonian, Latvian and Lithuanian Parakeet wins, so the right engine depends on the language. Parakeet is the faster of the two here: about 19 times faster than real time 3, against about 14 times for Whisper 4. Both are fast enough for dictation, so accuracy is the better reason to choose. The speed dataset covers every language.
Our measurement of the engines in Finnish
Lower is better in this table.
| Engine | Word error rate (WER %) | Source |
|---|---|---|
| Parakeet TDT v3 | 10.11 | Calculated 2 |
| Whisper large-v3-turbo | 7.73 | Calculated 1 |
The transcripts behind each row are on the on-device benchmark page.
Published scores point the same way
OpenAI's paper gives 9.7 percent for the older Whisper large-v2 on FLEURS Finnish 5, in line with our Whisper result. NVIDIA reports 13.21 percent for Parakeet on the same test 6, higher than our 10.11 percent. Whisper ahead of Parakeet therefore shows up in both our numbers and the published ones.
| System | Kind | Test set | Speech | Word error rate (WER %) | Source |
|---|---|---|---|---|---|
| Whisper large-v3-turbo | Open model | FLEURS | Clean read speech | 7.73 | Calculated 1 |
| Whisper large-v2 | Open model | FLEURS | Clean read speech | 9.7 | Radford et al., 2022 5 |
| Parakeet TDT 0.6B v3 | Open model | FLEURS | Clean read speech | 10.11 | Calculated 2 |
| NVIDIA Parakeet TDT 0.6B v3 | Open model | FLEURS | Clean read speech | 13.21 | NVIDIA, 2025 6 |
All published rows are on the published accuracy page.
Finnish ranks eleventh of 25 languages
Taking the best engine in each language, Finnish is eleventh of the 25 languages in the benchmark. It sits just behind Polish (7.70 percent 7), which scores almost the same, and ahead of Swedish (9.95 percent 8). Estonian, a close relative, is much worse at 14.17 percent 9. Finnish spelling follows pronunciation closely, which helps, though the benchmark does not test that reason.
Formatting costs about two points
Counting punctuation and capital letters as a reader sees them, Whisper's text differed from the written sentences by 9.88 errors per 100 tokens 10, and Parakeet's by 12.10 11. Commas and capitals are where the extra errors come from, so expect to fix punctuation as well as words.
What to expect when you dictate in Finnish
Expect to correct about 1 word in 13 with Whisper in a quiet room, which is good enough to dictate most of a draft and edit lightly. Parakeet leaves closer to 1 word in 10. Finnish builds long words from many endings, so names and compound words are the likely trouble spots, and a wrong ending counts as a whole wrong word. Add punctuation by hand or by voice command, and proofread. We found no published fact that confirms Finnish in a built-in Mac or Windows dictation list, so check your system's language list, and read our overview of dictation languages.
What these numbers do not show
The recordings are people reading prepared sentences in a quiet room, scored on one Mac with 100 recordings. Spontaneous speech, dialects, noise and specialist vocabulary are not in the test, and we did not measure Apple's model or any cloud service ourselves. Real dictation will usually score worse than read speech. For the measure itself, see the guide on how accurate speech recognition is.
Our guide to European languages in speech to text helps you choose an engine, and the languages overview covers other languages. For comparison, read the same question for Estonian, Swedish and Polish.
Evidence
Every number on this page, with where it comes from. Open a fact to see the exact words in its source.
- 1.
- 2.
- 3.
- 4.
- 5.
- 6.
- 7.
- 8.
- 9.
- 10.
- 11.
The data behind these numbers
- Speech Recognition Accuracy: Word Error Rates7 of this page's figures, 673 data points in all
- On-Device Speech Recognition Benchmark: 25 Languages5 of this page's figures, 56 data points in all
- Speech Recognition Speed: Real-Time Factor Data2 of this page's figures, 66 data points in all
- Dictation Punctuation and Capitalization Accuracy2 of this page's figures, 56 data points in all
Go deeper
Related questions
- Does dictation work for both Brazilian and European Portuguese?
- Does speech recognition understand Swiss German?
- How accurate is speech recognition in Bulgarian?
- How accurate is speech recognition in Croatian?
- How accurate is speech recognition in Czech?
- How accurate is speech recognition in Danish?
- How accurate is speech recognition in Dutch?
- How accurate is speech recognition in English?