Google has introduced two artificial intelligence (AI) releases – Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking – engineered to advance real-time logical deduction, power enterprise voice bots and deliver more natural spoken interactions. The updated releases are crafted to give software engineers and commercial enterprises the core components required to deploy dependable, deployment-ready voice systems.
Google also says that the technology enhances verbal interactions across the Gemini consumer app, Google Workspace and Search, enabling people to navigate multi-step challenges entirely through speech.
Gemini 3.8 Live has been optimised for high-volume deployment and operational affordability, and this model merges natural spoken conversation with visual contextual awareness. Gemini 3.8 Live Extended Thinking is tailored for sophisticated workflows requiring higher machine intellect and sequential multi-step problem-solving. it ranked first in agentic workflow resolution, recording 68.6% on the tau-Voice benchmark and 35.1% on Sierra’s banking-specific tau-Voice evaluation While Gemini 3.8 Live Extended Thinking secured the top spot worldwide on the Artificial Analysis Speech to Speech Quality Index with an 82.6 rating. The higher-tier model achieved a 97.7% score on the Big Bench Audio benchmark, as per Google.
To prevent uncomfortable dead air during computationally demanding operations, the system deploys instinctive verbal acknowledgments, such as saying “Let me check that…”, and provides running spoken commentary to update users on multi-phase processes as they unfold, according to Google. Google has segregated the rollout into two targeted configurations based on operational requirements. Both models recorded notable scores across established voice and audio evaluation metrics while preserving an affordable cost structure relative to rival leading-edge models. Get the latest technology news and updates. Download the TOI App.
Gemini 3.8 Live Model Enhancements
The Gemini 3.8 Live model offers rapid processing of incoming visual feeds, enhancing the relevance and richness of spoken responses. It supports 97 languages, enabling seamless language switching during conversations. Additionally, the software can execute background tasks and API requests without interrupting the dialogue, allowing for real-time validation of user commands while maintaining an ongoing conversation.
For more complex tasks, the Gemini 3.8 Live Extended Thinking feature can analyze underlying logic in real-time while communicating verbally.