Sources
Google shipped Gemini 3.8 Live and 3.8 Live Extended Thinking with 97 languages and async tool calls mid-speech
The two speech-to-speech models went live across the Gemini API, AI Studio, Gemini Enterprise private preview and consumer Search Live on 2026-09-15. Extended Thinking takes #1 on Artificial Analysis' Speech to Speech Quality Index at 82.6, 68.6% on τ-Voice agentic tasks, 35.1% on Sierra's banking benchmark and 97.7% on Big Bench Audio, while the standard variant sits second in Speech Agent Arena. The builder-relevant detail is background tool execution that does not interrupt the dialogue, plus early verbal cues like "Let me check that..." to cover latency, and AINews puts pricing at $0.84/hour of input audio standard and $3.50/hour for Extended Thinking.
Source
↳ Follow the thread