It outperforms Gemini 2.0 and Whisper v3, setting new benchmarks in accuracy. Leading in English, Spanish, Italian, and more, it supports 99 languages, speaker diarization, character-level timestamps, and non-speech events like laughter.