Google has upgraded its Gemini Live and Search Live platforms to the advanced Gemini 2.5 Flash Native Audio model. This new version enhances conversational interactions, significantly improving voice agent performance. Key features include the ability to manage complex requests, retain conversation context, and seamlessly interact with external sources. The Gemini 2.5 model greatly outperforms its predecessor, the 9-25 version, as well as OpenAI’s gpt-realtime model, achieving a 71.5% score in the ComplexFuncBench Audio benchmark. This upgrade boosts user satisfaction by allowing Gemini to execute multi-step tasks independently, reducing the need for human intervention. Developers can access the updated model via Google AI Studio, Vertex AI, and the Gemini API, enhancing the capabilities of Live Voice Agents. The rollout includes support for Android users, ensuring a richer conversational experience. Overall, the Gemini 2.5 Flash Native Audio model represents a significant advancement in AI-driven interaction quality.
Source link
Share
Read more