Google Launches Gemini 3.8 Live and Extended Thinking Audio Models

Google announced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking on Sep 15, 2026, introducing two live dialogue models designed for voice interactions and visual grounding. Gemini 3.8 Live processes camera feeds and visual inputs in near real-time, executing background API and tool calls while maintaining conversational flow, and automatically detects and transitions between 97 supported languages mid-conversation. The companion model, Gemini 3.8 Live Extended Thinking, targets multi-step workflows. It reasons and speaks simultaneously, using verbal cues like "Let me check that…" and live progress narration to guide users while asynchronous tasks run in the background.

Regarding costs, Google did not disclose specific per-token or per-minute dollar prices in the release post. The company stated that Gemini 3.8 Live is built for scale and cost efficiency, while Gemini 3.8 Live Extended Thinking is maintained at a competitive price point compared to other frontier models.

Google reports that across evaluation suites, Gemini 3.8 Live Extended Thinking secured the #1 overall spot on the Artificial Analysis Speech to Speech Quality Index with a score of 82.6. For agentic task completion, it marked 68.6% on τ-Voice and 35.1% on Sierra's τ-Voice-banking benchmark, alongside a 97.7% score on Big Bench Audio. Gemini 3.8 Live secured second place in the Speech Agent Arena. Both models were tested on ServiceNow's EVA-Bench, running on the Live API on Gemini Enterprise Agent Platform.

Both models roll out starting on the announcement date. Developers can access them through the Gemini API and Google AI Studio, alongside platform integrations such as Agora, Fishjam, LangChain, LiveKit, Pipecat, Vercel, and Vision Agents. Enterprises have access in private preview through Gemini Enterprise, with availability coming to Gemini Enterprise for Customer Experience. For general access, Gemini 3.8 Live is available in Search Live, while Gemini 3.8 Live Extended Thinking is in Gemini Live, with Workspace support rolling out to Docs, Gmail, and Keep. Audio output from both models is watermarked using SynthID.