Bringing Real-Time Reasoning and Task Execution to Enterprise Voice Agents Google’s latest voice models are designed to make AI agents more capable of reasoning, using tools and completing complex tasks while maintaining a natural conversation.
- Two new models: Gemini 3.8 Live targets low-latency, high-volume voice interactions, while Gemini 3.8 Live Extended Thinking is designed for more complex, multi-step tasks.
- Real-time reasoning: The models can reason during live voice interactions rather than simply transcribing speech and generating a response.
- Background task execution: Extended Thinking can execute asynchronous tools in the background while continuing the conversation, reducing the awkward pauses traditionally associated with voice agents.
- Enterprise use cases: Google specifically highlights applications such as customer-service triage, technical support, diagnostics, data retrieval and other multi-step workflows.
- Performance: Gemini 3.8 Live Extended Thinking reports 82.6 on Artificial Analysis’ Speech-to-Speech Quality Index, 97.7% on BigBench Audio, and 68.6% on the τ-Voice agentic task benchmark. These figures are reported by Google and should be treated as vendor-reported benchmarks.
- Availability: The models are available through the Gemini API and Google AI Studio, with enterprise availability also being expanded through Google’s ecosystem.
Source: Google DeepMind: Introducing Gemini 3.8 Live and 3.8 Live Extended Thinking