From the source
Google introduced Gemini 3.5 Transcribe, a speech-to-text model that converts raw audio into formatted text, handling background noise, jargon, and filler words.
It achieves a 4.0% word error rate for streaming and 2.6% for non-streaming, supports over 85 languages, and can identify up to three speakers.
The model is in public preview for developers via the Gemini API and for enterprises via the Gemini Enterprise Agent Platform.