How accurate is speech recognition in Greek?

Greek is usable with Whisper and poor with Parakeet. In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo had a word error rate of 11.96 percent ¹ on 100 Greek recordings from the FLEURS test set, while Parakeet TDT 0.6B v3 had 31.66 percent ². We have no Apple SpeechTranscriber result for Greek.

Whisper makes about one third of Parakeet's errors in Greek

This is the opposite of most languages in the benchmark, where Parakeet is level with or ahead of Whisper. Whisper is slower, at about 11 times faster than real time ³, against about 19 times for Parakeet ⁴. Both are fast enough for dictation, so speed should not decide it. The speed dataset covers every language.

Our measurement of the engines in Greek

Lower is better in this table.

The transcripts behind each row are on the on-device benchmark page.

Published results agree on Whisper but not on Parakeet

OpenAI's paper gives 12.5 percent for the older Whisper large-v2 ⁵, close to our 11.96 percent for the turbo model. NVIDIA reports 20.70 percent for Parakeet on FLEURS Greek ⁶, about eleven points better than our run. Even NVIDIA's figure is well behind Whisper, so the conclusion holds. Why our Parakeet figure is worse we cannot say from this data.

All published rows are on the published accuracy page.

Greek ranks seventeenth of 25 languages

Taking the best engine in each language, which for Greek is Whisper, Greek is seventeenth of the 25 in the benchmark. It trails Czech (10.44 percent with Parakeet ⁷) and Croatian (11.34 percent ⁸) only slightly, and is ahead of Bulgarian (13.42 percent ⁹). With the wrong engine it would fall to near the bottom of the field.

Formatting costs Whisper four points in Greek

Counting punctuation and capital letters as a reader sees them, Whisper's text differed from the written sentences by 16.10 errors per 100 tokens ¹⁰, and Parakeet's by 34.83 ¹¹. Whisper is still much better, but you will fix more commas, accents and capitals than the word error rate suggests.

What to expect when you dictate in Greek

In a quiet room with Whisper, expect roughly 1 wrong word in 8, so proofread every dictated passage. With Parakeet, expect nearly 1 in 3, which is not worth using for Greek. Errors are most likely on names, numbers, Latin-script words inside Greek text and the accent marks on words. If your dictation app lets you choose the engine, pick Whisper for Greek and keep Parakeet for other languages. Speak in full phrases and keep a correction list if your app has one. We found no published fact that confirms Greek as a built-in Mac or Windows dictation language, so check your system's language list.

Greek is the clearest case for choosing the engine by language

Most languages in the benchmark have two engines of similar quality, so the choice hardly matters. Greek does not. An app that uses one engine for every language will serve Greek well only if that engine is Whisper. Check which engine your dictation app uses for Greek before you rely on it, and test it with a page of your usual text. The published accuracy page shows how other models compare on similar tests.

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, noise, dialects and domain vocabulary are not in the test, and we did not measure Apple's model. 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 Bulgarian, Maltese and Croatian.