Google DeepMind Gemini 3.8 Live voice models

Google DeepMind released two real-time voice models on September 15, 2026: Gemini 3.8 Live, positioned as cost-efficient and scalable, and Gemini 3.8 Live Extended Thinking, targeting complex multi-step reasoning. Both models process real-time visual input and detect language automatically across 97 languages mid-conversation. Google DeepMind states that Extended Thinking ranked first on Artificial Analysis' Speech to Speech Quality Index with a score of 82.6 and leads agentic task completion benchmarks, while Gemini 3.8 Live placed second in the Speech Agent Arena. Extended Thinking reasons and speaks simultaneously, using verbal cues like 'Let me check that…' to maintain conversational flow, and executes tools and API calls in the background while continuing the conversation. All audio output carries a SynthID watermark, which Google DeepMind describes as woven directly into the audio output to keep AI-generated content detectable and help prevent misinformation.