Gemini 3.8 Live can keep talking while it works in the background, and switches between 97 languages
Google's new Gemini 3.8 Live voice models can see through your camera, run tasks in the background without pausing the conversation and narrate their progress. The Extended Thinking version tops a major speech-to-speech ranking.
By ScoviAI Team
Key points
- Two models: Gemini 3.8 Live (fast, cheaper) and Gemini 3.8 Live Extended Thinking (for complex tasks)
- Available in the Gemini API, Google AI Studio, Google Workspace, Search Live and the Gemini app
- Switches automatically between 97 languages mid-conversation and runs tool calls in the background
- Developer pricing: $0.005 per minute of audio input and $0.018 per minute of audio output
Voice assistants have long had an awkward habit: ask them to do something complicated and the conversation stops while they work. Google's new Gemini 3.8 Live models are designed to keep talking.
Google is releasing two versions:
| Model | Built for |
|---|---|
| Gemini 3.8 Live | Fast, cost-efficient conversations with real-time vision |
| Gemini 3.8 Live Extended Thinking | Harder, multi-step tasks with deeper reasoning |
What's new
It keeps the conversation going. Gemini 3.8 Live runs tools and API calls in the background, so it can acknowledge your request and keep chatting while the task finishes. The Extended Thinking version goes further: it reasons and speaks at the same time, uses natural cues like "Let me check that…" and narrates progress as a multi-step job runs.
It sees what you see. The model processes visual input in near real time, so it can use your camera's view as context. Google's demos include guiding a new employee through onboarding and playing chess.
It switches languages on its own. It detects and moves between 97 supported languages mid-conversation.
How good is it?
Google cites outside benchmarks. Gemini 3.8 Live Extended Thinking took the top spot on Artificial Analysis' Speech to Speech Quality Index with a score of 82.6, and scored 68.6% on τ-Voice, a test of how well voice agents complete tasks. The faster Gemini 3.8 Live came second in the Speech Agent Arena, a ranking based on user preferences. These figures come from Google's announcement.
Where you can use it
- Gemini app: ask for a Daily Brief, work through your inbox by voice or hand off to-dos.
- Google Workspace: Docs Live, Gmail Live and Keep Live.
- Search Live: step-by-step troubleshooting help.
- For developers: the Gemini API and Google AI Studio, through the Live API.
What it costs developers
Google lists $0.005 per minute of audio input and $0.018 per minute of audio output for the Live API. Voice platforms including Agora, LangChain, LiveKit, Pipecat and Vercel already support it.
Why it matters
Customer service lines, in-car assistants and accessibility tools all depend on voice AI that doesn't stall. A model that can keep a natural conversation going while it books, looks up or builds something in the background is a real step toward voice agents people will tolerate. Google's camera-based help for blind users, Guided Vision, runs in the same Gemini Live experience.
Sources
ScoviAI Team is the ScoviAI newsroom. We write every article ourselves from the original announcements and link our sources. Spotted a mistake? Tell us and we'll fix it.
ADVERTISEMENT
