Gemini 3.8 Live and 3.8 Live Extended Thinking (blog.google)

Article summary

Google has introduced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, its most advanced live dialogue models yet, which offer improved intelligence and parallel reasoning for more intuitive voice interactions. These models can handle complex tasks, real-time visual context, and background task execution without interrupting conversations. The models are available through the Gemini API, Google Workspace, and the Gemini app. They provide fast, fluid conversations with real-time visual and language support, and can execute tools and API calls in the background while continuing the conversation.

Main themes

• AI model advancements
• voice interaction
• conversational intelligence
• model usability
• performance vs usability
• AI strategy
• transparency and control
• error handling and robustness
• human-AI interaction

What commenters say

• Some commenters believe that Gemini's ability to play chess without using a guided decoding technique is a significant achievement, while others think it's not a useful feature.
• There is a debate about the value of models like Claude and Astra providing detailed, caveated responses versus Gemini's more straightforward answers.
• A few commenters have noted that Gemini's speed and simplicity make it preferable for simple tasks, but others may prefer more verbose models like Astra for complex tasks.
• Some users have experienced issues with Gemini hallucinating or providing inaccurate information, even on its Pro/Thinking model.
• There is a discussion about the trade-offs between model performance, usability, and transparency, with some commenters preferring models that provide more context and others valuing simplicity and speed.
• A few commenters have suggested using Gemini as a translator or interface to other models, allowing users to interact with more complex models in a simpler way.
• The release of new models has sparked a debate about Google's strategy and priorities in the AI space, with some commenters speculating about the company's goals and challenges.
• Some users have expressed frustration with the lack of transparency and control over which model they are interacting with, particularly with Astra.
• There is a concern that the pursuit of state-of-the-art performance may lead to models that are less usable or more prone to errors, and that a balance between performance and usability is necessary.