Google DeepMind launches Gemini 3.8 Live and 3.8 Live Extended Thinking for advanced voice interactions

Google DeepMind introduced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, positioning them as the company’s most advanced live dialogue models for voice-based interactions. The models emphasize faster, more fluid conversations, real-time visual context, and background task execution while maintaining an ongoing dialogue.

Gemini 3.8 Live is optimized for scale and cost efficiency, combining conversational intelligence with fluid dialogue and visual grounding. It supports near real-time processing of visual inputs and automatically detects and transitions between 97 supported languages mid-conversation.

Gemini 3.8 Live Extended Thinking is built for high-complexity tasks requiring multi-step reasoning. It leads in agentic task completion with 68.6% on τ-Voice and 35.1% on Sierra’s τ-Voice-banking benchmark, and scores 97.7% on Big Bench Audio for reasoning capabilities. It also ranks first on Artificial Analysis’ Speech to Speech Quality Index with a score of 82.6.

Both models are available today via the Gemini API, Google Workspace, and the Gemini app. They enable background execution of tools and API calls while continuing the conversation, allowing users to acknowledge requests and keep chatting as tasks complete.

Gemini 3.8 Live has demonstrated high user preference, securing second place in the Speech Agent Arena. On ServiceNow’s EVA-Bench, the models push the Pareto Frontier for complex workflows by balancing accuracy with conversational quality.