- Google officially announced “Gemini 3.8 Live,” a latest AI model specialized for voice interaction, and “Gemini 3.8 Live Extended Thinking,” which features more advanced reasoning capabilities.
- “Gemini 3.8 Live” is a voice-interaction-optimized AI model designed with an emphasis on scale and cost-efficiency, combining conversational intelligence, smooth dialogue, and responses based on visual information.
- “Gemini 3.8 Live Extended Thinking” is a voice-interaction-optimized AI model enhanced with stronger reasoning and multi-step processing capabilities to handle more complex tasks.
Just after announcing “Gemini 3.8 Flash” and “Gemini 3.8 Flash Cyber”, Google officially announced on Tuesday, September 15, 2026 (local time), its latest voice-interaction-focused AI models: “Gemini 3.8 Live” and the more advanced reasoning-capable “Gemini 3.8 Live Extended Thinking.”
Both “Gemini 3.8 Live” and “Gemini 3.8 Live Extended Thinking” aim to make voice interactions more natural, smooth, and intelligent, supporting complex reasoning, real-time visual awareness, and background task execution without interrupting the conversation. Rollout has already begun through the “Gemini API,” “Google Workspace,” and the “Gemini” app, among other channels.
“Gemini 3.8 Live”
“Gemini 3.8 Live” is a voice-interaction-focused AI model designed with an emphasis on scale and cost efficiency, combining conversational intelligence, smooth dialogue, and context-aware responses based on visual information.
It can process visual inputs from cameras and other sources in near real-time to add context to conversations, and automatically detect or switch between 97 supported languages mid-conversation. Additionally, because it can execute tools and API calls in the background while keeping the conversation going, it can acknowledge a request and proceed with the work without making the user wait for processing to complete.
“Gemini 3.8 Live” achieved 2nd place in user ratings on the “Speech Agent Arena,” which measures voice interaction performance, and is being deployed as a cost-efficient model for developers and enterprises while maintaining high performance.
Rollout for “Gemini 3.8 Live” has begun through the “Gemini API” and “Google AI Studio” for developers, via private preview of “Gemini Enterprise” for businesses, and through “Search Live” for general users.
“Gemini 3.8 Live Extended Thinking”
“Gemini 3.8 Live Extended Thinking” is a voice-interaction-focused AI model with enhanced reasoning and multi-step processing capabilities to handle more complex tasks.
Designed to reason and speak simultaneously, it handles complex tasks without breaking the flow of conversation. It can naturally acknowledge responses with phrases like “Let me check that for you,” and explain the progress of multi-step background tasks conversationally as they unfold.
“Gemini 3.8 Live Extended Thinking” demonstrated high reasoning performance by taking 1st place overall (82.6 points) in Artificial Analysis’s “Speech to Speech Quality Index” for voice interaction quality, scoring 68.6% on “τ-Voice” for agent task-execution capabilities, and 35.1% on the financial institution benchmark “τ-Voice-banking.”
In “Google Workspace,” voice interaction features such as “Docs Live,” “Gmail Live,” and “Keep Live” make it possible to handle complex tasks using only voice, while search services utilize it via “Search Live” for features like real-time step-by-step guidance.
Rollout for “Gemini 3.8 Live Extended Thinking” has begun for developers via the “Gemini API” and “Google AI Studio,” through private preview of “Gemini Enterprise” for businesses, and will expand in the future to “Gemini Enterprise for Customer Experience” and corporate plans for “Google Workspace.” For general users, it is rolling out through the “Gemini” app, “Workspace in Google Docs” for “Google AI Pro” and “Google AI Ultra” members, and “Gmail Live” and “Keep Live” for “Google AI Plus,” “Google AI Pro,” and “Google AI Ultra” members.
Source: Google (https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-live-gemini-3-8-live-extended-thinking/)