Google cuts speech-to-text latency 70 percent with Gemini 3.5 Transcribe
Source published ·Modelwire updated
Original coverage: The Decoder ↗·How Modelwire adds context
The development
Google's Gemini 3.5 Transcribe represents a meaningful step forward in production speech recognition, combining multilingual coverage (85 languages) with real-time error correction and a 4.0 percent word error rate in streaming mode. The 70 percent latency reduction versus Chirp 3 signals Google's focus on practical deployment speed. More strategically, the model's integration with function calling enables seamless handoff to other Gemini models, positioning speech as a first-class input modality within Google's broader AI stack. This matters for developers building voice-first applications and for enterprises evaluating speech pipelines where latency and accuracy directly impact user experience.
Modelwire’s AI-generated summary of coverage from The Decoder.
Modelwire analysis
Skeptical readOur AI-generated reading of the wider context and the next developments to watch.
Google hasn't disclosed whether the 4.0 percent word error rate applies uniformly across all 85 languages or concentrates in high-resource ones like English. The latency comparison is only against Chirp 3 (Google's own prior model), not against competing services like OpenAI Whisper or commercial alternatives, making the 70 percent improvement claim difficult to contextualize.
This is largely disconnected from recent activity in the broader speech recognition space, which we haven't covered at Modelwire. The story belongs to the category of incremental model releases where vendors report internal benchmarks without independent validation. The function calling integration is a tighter coupling story (speech as input to Gemini's broader agent capabilities), but without prior coverage of similar modality-stacking moves, we can't yet assess whether this represents a meaningful architectural shift or standard product bundling.
If Google publishes per-language error rates and they show WER above 8 percent for low-resource languages, the '85 languages' claim becomes marketing noise. Also track whether competing platforms (Anthropic, OpenAI) announce comparable multilingual speech models with published cross-platform benchmarks within the next six months. If they don't, Google may have a genuine capability gap; if they do quickly, this was incremental.
This interpretation is generated from the summary above and available source metadata. Our methodology · Report an error
MentionsGoogle · Gemini 3.5 Transcribe · Chirp 3
How this coverage is produced
Modelwire uses AI to generate summaries and context from source headlines, snippets, and selected archive coverage. Automated checks do not verify every claim, and items are not routinely reviewed by a person before publication. Zacaria Solis operates the site. Read the linked source for the full evidence and report errors through our corrections process.
Modelwire summarizes, we don’t republish. The Decoder originally reported this story as “Google's Gemini 3.5 Transcribe turns speech to text in 85 languages while auto-correcting your verbal stumbles”. The full content lives on the-decoder.com. If you’re a publisher and want a different summarization policy for your work, see our takedown page.