Gemini is getting a redesign and an even smarter new model

Google is announcing some big upgrades for Gemini at Google I/O, including a new design language for the app and a more intelligent 3.5 Flash model.

The company calls the new design language for Gemini “Neural Expressive,” which it says features “fluid animations, vibrant colors, new typography and haptic feedback.” It features a big “Ask Gemini” bar that’s easy to tap to start a conversation, and you’ll also see a slight gradient blue from the bottom — gradients, after all, are kind of Google’s thing right now.

Along with the new design language, Google is baking Gemini Live, which lets you have a back-and-forth voice conversation with Gemini, right into the core Gemini chat flow. Just press the Gemini Live button on the right side of the Ask Gemini bar to swap into Live mode, no separate screen required. And Google says that Gemini will create “tailored responses” that include things like images, timelines, and videos instead of just responding with a bunch of text.

The Neural Expressive design language is starting to roll out on Tuesday on Android, iOS, and the web.

As for its Gemini 3.5 Flash model, Google is touting its agentic and coding capabilities, calling it the company’s “strongest” model yet on those fronts. It says the model’s “balance of speed and performance” makes it a good choice for “tackling long-horizon agentic tasks.”

Gemini 3.5 Flash is becoming the default model in Gemini and in AI Mode starting today. Gemini 3.5 Flash is the model behind the new Gemini Spark AI agent for consumers, which was also announced today. And Google is promising that it’s working on Gemini 3.5 Pro, which it plans to roll out in June.