AIGoogle releases Gemini 3.8 Live and Extended Thinking for real-time voice AI
Google launched Gemini 3.8 Live for fluid conversational voice with visual grounding and tool use, plus Extended Thinking for reasoning and speaking simultaneously. Models top benchmarks including the Artificial Analysis Speech-to-Speech Quality Index at approximately 82.6.
TLDR
Voice is seen as the next key interface for AI; these models reduce friction for complex conversations and background tasks while reducing awkward pauses. Strong benchmarks, competitive pricing, and integration into Google products make them immediately usable across Gemini app, Search, and Workspace. The advances fuel industry discussion of voice agents moving from demos to production tools.
Combined views
—
2 Sources, first seen 2d ago
Google releases Gemini 3.8 Live and Extended Thinking for real-time voice AI
Google launched Gemini 3.8 Live for fluid conversational voice with visual grounding and tool use, plus Extended Thinking for reasoning and speaking simultaneously. Models top benchmarks including the Artificial Analysis Speech-to-Speech Quality Index at approximately 82.6.
TLDR
Voice is seen as the next key interface for AI; these models reduce friction for complex conversations and background tasks while reducing awkward pauses. Strong benchmarks, competitive pricing, and integration into Google products make them immediately usable across Gemini app, Search, and Workspace. The advances fuel industry discussion of voice agents moving from demos to production tools.