Google DeepMind released two conversational audio models capable of real-time dialogue and background processing for consumers and developers. Gemini 3.8 Live is optimized for scale and cost efficiency, whereas Gemini 3.8 Live Extended Thinking provides higher precision and multi-step reasoning for complex tasks. Both versions feature automatic detection for 97 languages, real-time visual understanding, and background tool calling to maintain conversation flow.
The rollout includes Search Live and the Gemini app for general users, with a public preview for developers in AI Studio. Gemini 3.8 Live Extended Thinking is also available for Workspace subscribers in Gmail, Google Docs, and Google Keep. API pricing is $0.005 per minute for input and $0.018 per minute for output. The models posted an 82.6 score on the Artificial Analysis Quality Index and a 35.1 agentic task completion score in tau-banking, while all generated audio is watermarked using SynthID.
Key sources
- SOURCE@googledeepmind“The models talk, think, and handle tasks in the background without breaking your flow”x.com
- SUPPORT@googledeepmind“Automatic detection for 97 languages”x.com
- SUPPORT@google“built for scale and cost efficiency, combining conversational intelligence with fluid dialogue and visual grounding”x.com
- SUPPORT@google“reasons and speaks simultaneously, while maintaining an uninterrupted conversational flow”x.com
- SUPPORT@google“All audio generated by our AI products is watermarked with @GoogleDeepMind's SynthID”x.com
- SUPPORT@_philschmid“$0.005/min input and $0.018/min output”x.com
- SOURCE@testingcatalog“Gemini 3.8 Live + Gemini 3.8 Live Extended Thinking have appeared on the Google Cloud quota & metrics page 2 hours ago”x.com
- SOURCEmarketbrief.now