AIGoogle releases Gemini 3.8 Live and Extended Thinking with real-time voice advances
Google launched Gemini 3.8 Live with optimized speech-to-speech models for low-latency real-time voice agents. Extended Thinking variant adds multi-step reasoning, scoring 82.6 on Speech-to-Speech Quality Index benchmarks.
TLDR
Voice and real-time agents represent a hot frontier for practical AI applications including smart homes, search, and productivity apps. The performance leap and immediate leaderboard dominance make this newsworthy, with implications for on-device/cloud hybrid experiences and competition in conversational AI. The Extended Thinking variant's reasoning capabilities address a key capability gap.
Combined views
—
1 Source, first seen 1d ago
Google releases Gemini 3.8 Live and Extended Thinking with real-time voice advances
Google launched Gemini 3.8 Live with optimized speech-to-speech models for low-latency real-time voice agents. Extended Thinking variant adds multi-step reasoning, scoring 82.6 on Speech-to-Speech Quality Index benchmarks.
TLDR
Voice and real-time agents represent a hot frontier for practical AI applications including smart homes, search, and productivity apps. The performance leap and immediate leaderboard dominance make this newsworthy, with implications for on-device/cloud hybrid experiences and competition in conversational AI. The Extended Thinking variant's reasoning capabilities address a key capability gap.