Gemini 3.8 Live keeps talking while tools run in the background
Google released Gemini 3.8 Live and 3.8 Live Extended Thinking with background tool calls and live progress narration, rolling Search Live and developer channels from 15 September. The Decoder reports the same models on the Gemini API and AI Studio, a first-place 82.6 percent Artificial Analysis speech-to-speech score, more than 97 languages, and 0.005 dollars a minute in and 0.018 dollars a minute out against 0.05 dollars at OpenAI.
Artificial Intelligence··Night
Google ships Live with spoken background tools
Google’s 15 September post says 3.8 Live is built to run tool and API calls in the background without breaking the spoken turn. Extended Thinking, it says, keeps multi-step work inside the conversation with early cues such as an early spoken check and live progress narration. 3.8 Live is rolling out from today to developers in the Gemini API and Google AI Studio, to enterprises in a Gemini Enterprise private preview, and to everyone in Search Live. Extended Thinking starts on the same developer channels and in the Gemini Enterprise private preview; customer-experience and Workspace business seats are listed as coming soon. For everyone, Google lists Gemini Live, Docs for Google AI Pro and Ultra subscribers, and Gmail and Keep for all Google AI subscribers. That is the announced calendar, not a claim that every surface is generally available at once.[1]
The Decoder adds rank, languages and per-minute prices
The Decoder, at 15 September 18:23 UTC, reports Google DeepMind released Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking through the Gemini API and Google AI Studio. Live can call APIs in the background, process visual input, and keep speaking, with more than 97 languages. Extended Thinking ranks first on the Artificial Analysis Speech-to-Speech Leaderboard at 82.6 percent, ahead of OpenAI GPT-Live-1. Google’s timed audio list price is 0.005 dollars a minute in and 0.018 dollars a minute out, against 0.05 dollars a minute at OpenAI. The Decoder puts an hour at about 1.38 dollars versus at least 3 dollars. It reports OpenAI still sounds more natural with full duplex. Sample apps are on GitHub.[2]
Background tools on two reports
Google frames Live as talk that continues while tools run in the background; The Decoder fills the same 15 September models with an 82.6 percent rank and 0.005 dollars / 0.018 dollars per-minute prices. Both texts present Extended Thinking as an audio model that works while it speaks. Google lists Search Live, the Gemini API, AI Studio and Gemini Live; The Decoder adds 97 languages, GitHub samples and OpenAI’s 0.05 dollars a minute. Unbroken speech is the shared event.[1], [2]
