Google Launches Gemini 3.8 Live and Extended Thinking Voice Models

Google launched Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking on September 15, 2026: near real-time voice dialogue models with visual grounding, 97-language mid-conversation switching, and background tool use. Extended Thinking adds simultaneous reasoning and leads Google-cited speech and agentic benchmarks, with rollout across Gemini API, AI Studio, Search Live, Gemini Live, and Workspace Live surfaces.

Source:Google

Post

Google Launches Gemini 3.8 Live and Extended Thinking Voice Models

TL;DRAI · 60 sec read

⚐ CORRECTED

Pre-publish harden (THE-128): removed claims not in Google’s Sep 15, 2026 primary — token limits 131072/65536, ultra-low/fixed latency profile wording, March 2026 3.1 Flash Live migration aside, and incorrect Workspace tier mapping (Docs: AI Pro/Ultra; Gmail/Keep: all Google AI subscribers).

→ All corrections

Google introduced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking on September 15, 2026, describing them as its most advanced live dialogue models yet.

Gemini 3.8 Live is positioned for scale and cost efficiency. Google says the model combines conversational intelligence with fluid dialogue and visual grounding. It processes visual inputs in near real-time, automatically detects and transitions between 97 supported languages mid-conversation, and can execute tools and API calls in the background while continuing to talk. Google also reports a high user preference result that placed the model second in the Speech Agent Arena, and frames it as a capable, efficient option built for scale.

From The CircuitryThe Feed — live briefs across tech, all day.See what’s happening →

Gemini 3.8 Live Extended Thinking targets higher-complexity work. Google says the model reasons and speaks simultaneously, using early verbal cues such as “Let me check that…” and live progress narration so multi-step background tasks can continue without breaking conversational flow. On Artificial Analysis’ Speech to Speech Quality Index, Google reports a score of 82.6 and the #1 overall spot. Google also cites 68.6% on τ-Voice and 35.1% on Sierra’s τ-Voice-banking for agentic task completion, plus 97.7% on Big Bench Audio.

Rollout surfaces span developers, enterprises, and consumer Google products. According to Google, 3.8 Live is rolling out in the Gemini API and Google AI Studio for developers; in private preview in Gemini Enterprise for enterprises (with Gemini Enterprise for Customer Experience described as coming soon); and in Search Live for everyone. 3.8 Live Extended Thinking is likewise available in the Gemini API and Google AI Studio; in private preview in Gemini Enterprise (with broader enterprise and Workspace business availability described as coming soon); and for everyone in Gemini Live, with Google AI Pro and Ultra subscribers getting Workspace access in Docs, and all Google AI subscribers getting access in Gmail and Keep.

Google also points developers to the Live API ecosystem and safety controls. Platforms cited as enabling voice-driven interfaces via the Gemini Live API include Agora, Fishjam, LangChain, LiveKit, Pipecat, Vercel, and Vision Agents. Google says all audio generated by its AI products is watermarked with SynthID and points readers to the model card for safety and responsibility details.

Why this mattersAI · ~100 words

Tap a lens to see what this story means for you.

Reader-supported

DonateBuy me a coffee →Follow@thecircuitry_ →Follow@thecircuitry.to →

Reader-supported · The Brief

Liked this? The Brief brings you the whole day in tech, verified, every morning. Two minutes, free forever.

HELP US IMPROVE

Reader-supported

The Circuitry is a passion project I've always wanted to build, and I love the work behind it.

Running it costs real money. APIs, hosting, time. To keep improving the site and growing this into something useful for everyone, those costs have to be covered.

Any contribution is appreciated. If not, no pressure. Thanks for reading.

Buy me a coffee

MORE IN TECH

Microsoft AI Opens Six-Week Consultation on Humanist Code of Conduct for MAI Models

Microsoft AI released a draft Humanist AI Code of Conduct for MAI models on September 14, 2026, opening six weeks of public feedback. The company says the draft puts people first, keeps models interruptible and auditable, sets Absolute Constraints on weapons, offensive cyber, and mass manipulation, and will be revised later in 2026 to guide 2027 training — current models are not trained on it yet.

CISA Adds Cisco Email Gateway Flaw to KEV Catalog

CISA added CVE-2026-76461, a SQL injection flaw in the Cisco email security appliance, to its Known Exploited Vulnerabilities catalog on 2026-09-14. Federal agencies have until 2026-09-17 to apply vendor mitigations under BOD 26-04, with forensic triage required.

Apple Releases tvOS 27 for Apple TV 4K

Apple released tvOS 27 on September 14, 2026, bringing a redesigned Podcasts app, larger text options, customizable subtitles, a more responsive Control Center, and performance improvements to the Apple TV 4K. It drops support for the Apple TV HD and first-generation Apple TV 4K while delivering faster AirPlay, quicker app launches, and smoother animations on compatible models.