How accurate is speech recognition in Italian?

Italian is the most accurately recognized language in our benchmark: the best on-device engine makes fewer than 3 wrong words per 100 on read speech. In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 had a word error rate of 2.52 percent ¹, Whisper large-v3-turbo had 2.73 percent ², and Apple SpeechTranscriber had 4.15 percent ³. The test was 100 Italian recordings from the FLEURS set, all run on the device.

Apple's model is fast and good enough in Italian

Apple's built-in model is reasonably close to the open engines in Italian, and it is the fastest of the three by a wide margin: about 75 times faster than real time ⁴, against about 21 times for Parakeet ⁵ and about 17 times for Whisper ⁶. Parakeet and Whisper are within about a fifth of a point of each other, which is within noise on 100 recordings. The speed dataset covers every language.

Our measurement of all three engines in Italian

Lower is better in this table.

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

Published results depend heavily on the test set

On FLEURS Italian, the Voxtral paper reports 1.48 percent ⁷ for ElevenLabs Scribe, 2.02 percent ⁸ for GPT-4o mini Transcribe and 2.71 percent ⁹ for Whisper large-v3. NVIDIA reports 3.00 percent ¹⁰ for Parakeet.

Change the test set and the picture changes. On Multilingual LibriSpeech Italian, ElevenLabs Scribe had 12.38 percent ¹¹ and Parakeet had 10.08 percent ¹², several times worse than on FLEURS. Italian is therefore not always the easiest language: its low error rate holds for clean read sentences.

Showing 8 of 27 rows. See every row in the dataset.

All published rows are on the published accuracy page.

Italian ranks first of 25 languages by its best engine

Taking the best engine in each language, Italian has the lowest error rate of the 25 languages in the benchmark, just ahead of Spanish (2.53 percent with Whisper ¹⁷). English (3.92 percent ¹⁸) and German (4.84 percent ¹⁹) follow. The lead over Spanish is small enough that the two are best treated as tied. Italian spelling is close to pronunciation, which plausibly helps, though the benchmark does not test that.

What to expect when you dictate in Italian

Expect around 3 wrong words per 100 with Parakeet or Whisper in a quiet room, and somewhat more with Apple's built-in model. Because Apple's engine is close in accuracy and much faster, the built-in Mac dictation is a sensible first choice. Switch to Parakeet or Whisper if you see too many errors on names, numbers or technical terms. Expect harder audio, such as phone calls or crowded rooms, to cost far more than the choice of engine does.

Our guide to European languages in speech to text helps you choose an engine, and the languages overview covers other languages.

How to choose an engine for Italian

With error rates this low, the practical differences between engines are small for everyday use. Parakeet and Whisper give the fewest corrections, and Apple's model is the quickest and already built into the Mac. If you dictate long documents, take the lowest error rate. If you dictate short messages all day, take the fastest. Whichever you pick, speak a minute of your usual text into two engines and count the corrections: your accent, vocabulary and microphone will matter more than the small gaps in the table above. The published accuracy page shows how cloud services compare.

What these numbers do not show

The recordings are people reading prepared sentences in a quiet room, scored on one Mac. Regional accents, dialects, spontaneous speech, noise and domain vocabulary are not in the test, and 100 recordings is a small sample. For the measure itself, see the guide on how accurate speech recognition is. Compare with Spanish and French.