XNews

Google Gemini 3.8 Live and Extended Thinking launch

Google Gemini 3.8 Live and Extended Thinking launch

Google DeepMind announced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, new SOTA live audio models. The base version supports 97 languages with automatic switching, async tool calls, and near real-time visual understanding. Extended Thinking adds multi-step reasoning and planning for complex enterprise tasks, narrating progress during operations. Both models are available in Gemini Live and via the Gemini API.

Key insights

Gemini 3.8 Live targets high-volume, low-latency applications

The base Gemini 3.8 Live model is designed for cost-effective, large-scale deployment, supporting 97 languages with automatic detection and smooth transitions. It enables seamless multimodal real-time interaction, including visual and voice, while maintaining competitive operational costs. This positions it for everyday assistant and retrieval services requiring high concurrency and low latency.

Extended Thinking enhances complex task execution with real-time narration

Gemini 3.8 Live Extended Thinking focuses on enterprise-grade tasks, strengthening multi-step logical reasoning and planning. It synchronizes real-time thinking with speech, using phrases like 'let me check' as transitions while executing background tasks. The model supports asynchronous tool calls, allowing uninterrupted voice chat during API calls or data retrieval, and provides near real-time visual understanding for tasks like coding or answering questions based on visual input.

Source posts