XNews

Google DeepMind Launches Gemini 3.8 Live Audio Models

Google DeepMind Launches Gemini 3.8 Live Audio Models

Google DeepMind announced Gemini 3.8 Live and 3.8 Live Extended Thinking, their new state-of-the-art conversational AI models. The models support 97 languages, near real-time visual understanding, and background tool calling, and are available in the Gemini App and via the Gemini API.

Key insights

New SOTA Live Audio Models Released

Google DeepMind introduced Gemini 3.8 Live and 3.8 Live Extended Thinking, described as their best conversational AI yet. The 3.8 Live model supports 97 languages with seamless switching, async tool calls, and frontier-level performance. The Extended Thinking variant adds increased performance and precision for difficult tasks, narrating its progress to keep the conversation going.

Models Feature Upgraded Reasoning and Multimodal Capabilities

The models feature upgraded reasoning, near real-time visual understanding, and automatic language detection for 97 languages. A key capability is background tool calling, which allows the AI to perform tasks without interrupting the user's conversation flow, as demonstrated in a programming tutor use case.

Availability and Developer Access

The models are available for testing in Gemini Live within the Gemini App and for developers via the Gemini API through Google AI Studio. An in-person Gemini Audio experience event is scheduled for the following week in San Francisco, where the new models will be showcased.

Source posts