Google introduces Gemini 3.5 Flash , the first model in their next generation, built to combine deep intelligence with lightning-fast execution.
At the same time, a completely new design language called Neural Expressive is being rolled out globally. This means fluid animations and haptic feedback, but most importantly, it changes the way you get answers. Instead of large blocks of text, Gemini now designs interactive answers in real time with timelines, graphics and videos. At the same time, the microphone has been updated (Gemini Live), so you can speak at your own pace without being interrupted in the middle of a thought.

For the creatives, Gemini Omni is introduced. The model can mix text, images and video prompts and spit out finished videos in high quality. You can edit directly through a conversation – for example, add zoom or change the background. It is even possible to upload video of yourself and create a custom AI avatar that looks and sounds like you. The feature is rolling out now to AI Plus, Pro and Ultra subscribers.
Google launches two significant agent features that run in the background – even when your phone is locked or your computer is closed

Security is built into the design, so Spark always asks for permission before performing important actions like sending emails or spending money.
Google is also releasing an official macOS app , available for download starting today. This summer, Gemini Spark will be integrated into the app, allowing it to work with your local files directly on your computer. The Mac app will also get an advanced voice feature that can listen to your random thoughts, remove your "uhs," and turn them into precise text drafts right where your cursor is on the screen.
Quick specs: