In May 2026, Google delivered a major upgrade to Gemini 2.5 Flash Native Audio. The standout features are live speech-to-speech translation that preserves original tone, pace, and pitch β and Proactive Audio, which only responds when the AI is being addressed. Voice AI is learning to listen like a person, not a microphone.
Google launched Gemini 2.5 Flash-Lite to general availability on Vertex AI. It uses 20-30% fewer tokens than the standard Flash while maintaining reasoning, coding, and multimodal performance β and it now supports supervised fine-tuning (SFT). At the same time, Deep Research opened for free on the Flash model, and Native Audio got sharper. A significant update for cost-conscious teams and individuals alike.
In May 2026, Google significantly upgraded Gemini 2.5 Flash Native Audio. Sharper function calling, smoother conversation flow, and multi-speaker TTS β creating audio-based educational content is now possible with a few lines of code. An EdTech CEO shares real use cases and API implementation details.