Overview
Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking are Google’s native speech-to-speech voice dialogue models, announced September 15, 2026. They support real-time visual grounding, 97-language mid-conversation switching, and asynchronous background tool execution.
Variants
| Model | Focus | Benchmark |
|---|---|---|
| Gemini 3.8 Live | Scale, cost efficiency | #2 Speech Agent Arena |
| Gemini 3.8 Live Extended Thinking | Multi-step reasoning | #1 Artificial Analysis Speech-to-Speech (82.6) |
Availability
- Gemini API / Google AI Studio: 0.018/min output
- Consumer: Gemini Live, Gmail Live, Docs Live, Keep Live
- Enterprise: Private preview in Gemini Enterprise
- Integration partners: LiveKit, LangChain, Pipecat, Vercel