XNews

Google Launches Gemini 3.8 Live and Extended Thinking

Google Launches Gemini 3.8 Live and Extended Thinking

Google released two new live audio AI models: Gemini 3.8 Live for cost-efficient conversational AI, and Gemini 3.8 Live Extended Thinking for complex multi-step reasoning. Both support 97 languages, near real-time visual understanding, and background tool calling without disrupting conversation.

Key insights

Gemini 3.8 Live Offers High-Concurrency Conversational AI

The base Gemini 3.8 Live model is designed for large-scale deployment with high concurrency and low latency, targeting everyday assistants and retrieval services. It automatically detects and smoothly transitions across 97 languages, and supports real-time multimodal interaction including vision and voice.

Extended Thinking Version Adds Synchronized Real-Time Reasoning

Gemini 3.8 Live Extended Thinking focuses on enterprise-grade complex tasks, enhancing multi-step logical reasoning and planning. It synchronizes real-time thinking with speech, naturally saying phrases like 'let me check that' while executing background tasks and reporting progress, creating a phone-call-like interaction.

Background Tool Calling Maintains Conversational Flow

The models support asynchronous tool calls in the background, such as calling external APIs, retrieving data, or making reservations, while the front-end voice conversation continues uninterrupted. This enables true concurrent collaboration.

Near Real-Time Visual Understanding Capabilities

The models can see and understand visual input while chatting, allowing applications such as answering employee onboarding questions, playing chess by looking at the board, or generating React code from voice descriptions of sketches.

Source posts