Google Unveils Gemini 3.8 Live for Enhanced Voice Assistance

WHO:

9to5Google - Google announced the release of Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking.

MarkTechPost - Google DeepMind is associated with the audio generated by the models.

WHAT:

9to5Google - Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking are described as Google’s “most advanced live dialogue models yet.”

9to5Google - The models aim to make conversing with AI feel more intuitive and intelligent.

9to5Google - Gemini 3.8 Live Extended Thinking offers “increased intelligence and multi-step reasoning” for high-complexity tasks.

MarkTechPost - Extended Thinking ranks #1 on Artificial Analysis' Speech to Speech Quality Index with a score of 82.6.

MarkTechPost - The model scores 97.7% on Big Bench Audio.

MarkTechPost - Both models are available for $0.005/min for audio input.

WHERE:

MarkTechPost - The models are available in the Gemini API and Google AI Studio.

WHEN:

9to5Google - The models are beginning to roll out today.

MarkTechPost - All generated audio carries Google DeepMind's SynthID watermark.

WHY:

9to5Google - The models are designed to deliver increased intelligence for complex workflows while maintaining an uninterrupted conversational flow.

HOW:

9to5Google - The models execute tools and API calls in the background while the conversation continues.

9to5Google - Gemini 3.8 Live can process visual inputs in near real-time and can detect and transition between 97 supported languages mid-conversation.

(UPDATED September 16, 2026 02:00 UTC)

Google has just announced the release of its latest advancements, Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, which are touted as the company's most advanced live dialogue models yet. These models aim to enhance the conversational experience with AI, offering features like increased intelligence for complex tasks and the ability to process visual inputs in near real-time, while seamlessly transitioning between 97 languages. Both models are available for $0.005 per minute for audio input through the Gemini API and Google AI Studio, and they carry Google DeepMind's SynthID watermark for all generated audio.