What happened
On September 15, 2026, Google introduced two live dialogue models it calls its most advanced yet:
- Gemini 3.8 Live: built for scale and cost efficiency, combining conversational intelligence, fluid dialogue and visual grounding.
- Gemini 3.8 Live Extended Thinking: built for high-complexity tasks, with more intelligence and multi-step reasoning.
Google says the models handle interruptions better, switch languages mid-conversation and can explain their reasoning while solving a problem by voice.
Where it is available
- Developers: Gemini API and Google AI Studio (both models).
- Enterprises: private preview in Gemini Enterprise; coming to Gemini Enterprise for Customer Experience.
- Consumers: 3.8 Live in Search Live; Extended Thinking in Gemini Live and, for Google AI subscribers, in Gmail, Keep and Workspace Docs.
Why it matters
Voice agents for customer service have always hit two limits: latency and shallow reasoning. Splitting a cheap, fast model from one that "thinks" more lets you route each call to the right cost tier.
What to do now
- For contact centers, test 3.8 Live on FAQs and Extended Thinking only on multi-step cases.
- Measure real latency with your own audio and accents, not just the official demo.
- Review consent and call recording rules before putting a voice agent in production.