Gemini 3.5 Transcribe
by Google · Model · Voice & audio
Transcription model with verbatim and smart modes, speaker detection, word timestamps and automatic multilingual handling
AccessAPIPrice$2 in / $12 out per M tokens
Evidence
- Aug 30On Pixel 11 Pro dictation, a user found it simplifies phrasing, dropping clauses like 'I hesitated to check it' and changing meaning.
- Aug 27Gemini 3.5 Transcribe and a Live (streaming) variant released in the Gemini API; also powers Pixel dictation.
- Aug 27Reported as still more expensive and worse performing than ElevenLabs Scribe.
- Aug 27Real-time translator developer found it most accurate of tested models but too slow; Soniox STT v5 wins on latency.
- Aug 27Added to a multilingual STT benchmark site; it does not reach the frontier in either latency or accuracy there.