Fact · Speech recognition
32tokens per second of audio
Google's audio guide says Gemini models use 32 tokens for each second of audio, which is 1,920 tokens per minute.
Checked against its source on October 17, 2026
In the source's own words
“32 tokens per second of audio (1 minute = 1,920 tokens)”
Audio understanding (Gemini API), Google, Google AI for Developers, 2026, Technical details about audio.
Open the original- What it measures
- The technical details section of Google's audio guide, which covers audio passed to Gemini in a prompt. It also lists non-speech sounds such as birdsong and sirens as recognized input.
- What it does not show
- This rate describes general audio input in a prompt. The transcription model's pricing uses 25 tokens per second, so a per-minute cost worked out from this figure will not match the price list for that model.