Press "Enter" to skip to content

Posts tagged as “Gemini 3.8 Live”

Google Announces Gemini 3.8 Live and 3.8 Live Extended Thinking

September 16, 2026
Google Announces Gemini 3.8 Live and 3.8 Live Extended Thinking

Google has introduced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, marking a major leap in real-time, audio-to-audio artificial intelligence. These models are built specifically to handle fluid, back-and-forth voice interactions while simultaneously processing complex underlying logic.

Core Capabilities

  • Gemini 3.8 Live: Optimized for low-latency, natural dialogue. It allows users to speak organically with the AI, supporting real-time interruptions, dynamic tone adjustment, and fluid conversational flow without typical voice-assistant delays.
  • Gemini 3.8 Live Extended Thinking: Tailored for complex tasks that require higher background reasoning. By leveraging parallel processing, the model executes multi-step logic and problem-solving “behind the scenes” while maintaining an uninterrupted, natural spoken conversation.

Key Benchmarks & Features

  • Audio Intelligence: Achieves 97.7% on Big Bench Audio, establishing a new benchmark for spoken instruction comprehension and context retention.
  • Continuous Audio Stream: Operates natively in an audio-to-audio framework rather than converting voice to text and back, drastically reducing response latency.
  • Parallel Reasoning: Extended Thinking actively resolves edge cases, coding problems, or computational questions mid-conversation without pausing the verbal output.

These updates transform voice agents from simple command-execution tools into active collaborative partners. Developers can deploy Gemini 3.8 Live for live customer support, conversational tutoring, or real-time translation. Meanwhile, the Extended Thinking variant enables complex hands-free workflows—such as debugging code via voice or working through multi-variable strategy problems during a call—making voice interaction far more versatile across technical and professional fields.

You can read the full text of Google’s announcement at Blog.Google.