The Next Generation of Voice Interaction: Meet Gemini 3.8 Live and Extended Thinking
Google continues to tighten its grip on the field of voice AI with the launch of two new models: Gemini 3.8 Live and a dedicated Extended Thinking version. While the market is flooded with general models, Google has chosen this time to focus on a more fluid user experience, one that does not halt the flow of conversation even when the system is performing complex operations in the background.
Why does this matter?
The main problem with most voice agents to date has been "thought latency." The moment you asked the agent to perform a complex task, the conversation would stop until the answer was received. Gemini 3.8 Live changes the rules of the game: the model was designed so that the conversation continues to flow naturally, even while the agent is running external tools or processing heavy data behind the scenes [2, 4].
What's new under the hood?
The combination of the two new models offers capabilities we have not yet seen in Google's consumer products:
- •Extended Thinking: This is the highlight for developers. The model allows you to define the "depth of thought" required for the task. As the agent performs multi-step reasoning processes, it can continue to update the user on its progress, which creates a sense of transparency and control [2, 8].
- •Live Visual Input: The new models do not just "hear," but are capable of processing visual input in real-time, allowing them to respond to the user's environment in a more intelligent way [2].
- •Broad Support: Google reports support for more than 97 languages, making this technology accessible to a wide global audience [6].
Availability for developers and organizations
Google is not waiting – the models are available now for developers via the Gemini API and Google AI Studio [4]. For organizations, these capabilities are gradually being integrated into Gemini Enterprise, allowing for the implementation of smart agents within complex business workflows [4].
Summary and conclusions
The launch of the 3.8 series marks a shift from AI that only "answers questions" to AI that "manages processes." The model's ability to keep talking while it thinks or performs actions is critical for the adoption of AI agents in the real world. If you are a developer or business owner looking to improve your customer experience, now is the time to start experimenting with Extended Thinking in AI Studio and see how it changes the dynamics of interaction with your users.
Additional sources for further reading: