Google's Gemini 3.5 Transcribe slashes speech-to-text errors by 70% in 85 languages

google has launched gemini 3.5 transcribe, a speech-to-text model capable of recognizing over 85 languages in real-time. it has a word error rate of 4.0 percent for streaming and 2.6 percent for recorded audio, with 70 percent lower latency than its predecessor, chirp 3. the model auto-corrects verbal stumbles, strips filler words, and formats text automatically. it is available through two interfaces: the live api for real-time streaming and the interactions api for recorded audio with speaker attribution and timestamps. gemini 3.5 transcribe is live in google ai studio and on the gemini enterprise agent platform, and is already integrated into gboard for android and the gemini app on macos.