Google Releases Gemini 3.1 Flash Live AI Model, Making Conversations with AI More Human-like
Google has officially released its latest audio artificial intelligence (AI) model, Gemini 3.1 Flash Live. This model is designed to allow users to “chat” with AI in a more natural way, akin to speaking with a human.
In its official blog post, Google states that Gemini 3.1 Flash Live is the highest-quality audio and voice AI model it currently possesses.
Google explains that this model brings improvements in precision and lower latency. As a result, the generated voice responses become faster, more accurate, and feel natural.
Google claims that this setup can provide an intuitive interaction experience for developers, companies, and general users.
Google notes that the overall quality of Gemini 3.1 Flash Live has been enhanced to be more reliable in handling large-scale complex tasks, particularly in voice-based conversations.
This AI model has been tested using ComplexFunchBench Audio. It is an AI benchmark used to measure and test the model’s ability to understand complex instructions and perform “function calling” (sequentially calling APIs/tools).
The result is that the model achieved a score of 90.8%. In comparison, on the Audio MultiChallenge from Scale AI, with “thinking” mode active, the Gemini 3.1 Flash Live audio AI model achieved a total score of 36.1%.
This benchmark tests the ability to follow complex instructions and long-term reasoning, amidst distractions and pauses common in real-world audio conversations.
The model is also said to be more capable of adjusting responses according to the user’s condition. For example, when the user sounds confused or frustrated, Gemini 3.1 Flash Live will provide answers more suited to the situation.
Gemini 3.1 Flash Live itself is available on various Google services. The model is used in the Search Live and Gemini Live features for general users, and can be accessed by developers through the Gemini Live API in Google Studio.
For business needs, the model is also available through the Gemini Enterprise service for Customer Experience.