Google launches Gemini 3.8 Live voice AI models

Google Alphabet Inc. ↗ 0.78% on Tuesday released Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, a pair of voice-first AI models designed for real-time conversation, visual understanding, and background task execution. The company calls them its most advanced live dialogue models to date.blog+1

The launch positions Google to compete directly with OpenAI's GPT-Live-1 in the emerging voice agent market — at a fraction of the price. According to The Decoder, Gemini 3.8 Live Extended Thinking costs roughly $1.38 per hour of voice conversation, undercutting OpenAI's offering by roughly 13 times. On Artificial Analysis' Speech to Speech Quality Index, Extended Thinking scored 82.6, claiming the top spot, compared to GPT-Live-1's 79.4.autointerviewai+1

What the Models Do

Gemini 3.8 Live is built for scale and cost efficiency, processing visual inputs in near real time and automatically switching between 97 languages mid-conversation. It can execute tool calls and API requests in the background while continuing to talk, acknowledging a user's request verbally and picking the conversation back up once the task completes.unite+1

Extended Thinking adds deeper reasoning on top of that conversational layer. Google says the model reasons and speaks simultaneously, using verbal cues like "Let me check that…" to maintain natural flow while working through multi-step problems. It scored 68.6% on the τ-Voice agentic task benchmark and 97.7% on Big Bench Audio.blog+1

Demonstration videos show the models guiding employee onboarding with live visual context, converting hand-drawn sketches into functional React components via voice feedback, and coordinating multi-step restaurant bookings through asynchronous function calls.unite

Who Gets Access

Both models began rolling out Tuesday. Developers can access them through the Gemini API and Google AI Studio. Enterprises get a private preview in Gemini Enterprise, with customer experience support coming later. For consumers, Gemini 3.8 Live powers Google Search Live, while Extended Thinking is available in the Gemini Live app, in Docs for Google AI Pro and Ultra subscribers, and in Gmail and Keep for all Google AI subscribers.sqmagazine+1

Google named Agora, LiveKit, Pipecat, Vercel, and Salesforce ↘ 3.94% among early platform and enterprise partners building on the models' low-latency tool-calling capabilities. All audio output is watermarked with Google's SynthID system to flag AI-generated speech.blog+1

The Competitive Picture

The release sharpens the rivalry in voice AI. OpenAI's GPT-Live-1 retains an architectural advantage — true full-duplex conversation, listening while it speaks — but Google is betting that superior benchmarks and aggressive pricing will win over developers building production voice agents at scale. On ServiceNow's ↘ 0.38% EVA-Bench, Google says its models push the accuracy-versus-conversational-quality frontier for complex voice workflows.orcarouter+3