Google Releases Gemini 3.8 Live and Live Extended Thinking

Google introduced two live voice models — Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking: 97 languages with automatic mid-conversation switching, near real-time camera video, and background API call execution.

Compiled by Sergey KostenchukPublished 2026-09-15Updated 2026-09-18

2026-09-15 Research Google

🎙 Google Releases Gemini 3.8 Live and Live Extended Thinking

Google introduced two live voice models, gemini-3.8-live and gemini-3.8-live-extended-thinking. They speak 97 languages with automatic mid-conversation switching and understand camera video in near real-time, while executing tool and API calls in the background without interrupting the conversation.

🌍 According to Google, Extended Thinking ranked #1 on the Artificial Analysis Speech to Speech Quality Index (82.6 points). The API price — $0.75/$4.50 per 1M text tokens and $3/$12 for audio — sets a benchmark for the real-time voice agent segment.

👤 Developers can already try the models in Google AI Studio (a free tier is available) and connect them to their voice agents via the Gemini API. They will soon be available in Search Live, Gemini Live, and Workspace.

Source 1: https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-live-gemini-3-8-live-extended-thinking/

Sources