Google Launches Gemini 3.8 Live and 3.8 Live Extended Thinking for Real-Time Interaction

It is a multimodal model from Google DeepMind designed for real-time interaction via voice, image, and video, making it suitable for interactive applications such as live support and voice agents.

The Live version prioritizes fast responses for real-time interaction, while Extended Thinking spends more time reasoning before answering, improving quality on complex tasks at the cost of higher latency.

It is advisable to start with a limited pilot to measure latency, cost, and data-compliance fit, especially for government and regulated-sector use cases.