Official provider results
Gemini 3.5 Transcribe launch evaluation
Google’s August 26, 2026 launch post reports word error rates on an independent transcription leaderboard and FLEURS for pre-recorded and streaming transcription, plus a latency gain over Chirp 3.
Source snapshots refreshed
Five measurement lanesVOICE / S2S
Reasoning, task completion, conversation dynamics, experience, and latency stay separate.
Benchmark profile
- Scope
- independent-leaderboard WER (non-streaming and streaming), FLEURS multilingual WER, and time-to-final-transcript versus Chirp 3
- Primary metric
- Word error rate (lower is better)
- Owner
- Available evidence
- Exact figures published in the official launch post
What it measures
Voice experience
Is the exchange natural, robust, and responsive?
Latency
How long do responses, tool calls, and full tasks take?
Available results
- Independent WER leaderboard (pre-recorded)
- 2.6%
- Independent WER leaderboard (streaming)
- 4.0%
- FLEURS WER (pre-recorded)
- 5.04%
- FLEURS WER (streaming)
- 5.50%
- Time to final transcript
- 70% faster than Chirp 3
Provider-reported
Average across the independent leaderboard’s transcription set; lower is better.
Provider-reported
Real-time streaming mode.
Provider-reported
Multilingual FLEURS benchmark.
Provider-reported
Multilingual FLEURS benchmark, streaming mode.
Provider-reported
Relative latency claim versus Google’s previous transcription model.
Interpretation limit
These are provider-reported launch results against Google’s own prior model; the post gives no per-language breakdown or competitor rows.