Google launched Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, speech-to-speech models aimed at real-time conversational agents. The pair is available in the Gemini API, Google AI Studio, Search Live, and Gemini Enterprise, according to company and trade coverage dated September 15–17.
Gemini 3.8 Live is positioned as a cost-efficient conversational endpoint. Live Extended Thinking is built for multi-step reasoning that continues while the model keeps speaking. Reporting placed the family at the top of independent speech-to-speech leaderboards, with real-time vision and switching across dozens of languages.
Voice is now a product surface, not a demo. OpenAI’s GPT-Live-1 already offered full-duplex API access; Google’s answer emphasizes simultaneous reasoning and lower hourly cost than the premium OpenAI voice stack. That pricing gap matters for call-center, device, and in-car agents that stay on the line for minutes, not seconds.
The release also tightens Google’s loop between Search Live and the developer API. A model that can see, speak, and think without dropping the conversation is the missing piece for agents that control homes, cars, and customer queues.
Key takeaway Google is competing on live voice quality and cost at once — reasoning in the background without pausing the conversation.
Photo: Unsplash. Sources: AI Weekly, BitsMinds, donvitocodes.com, Sept. 15–17, 2026.
