Google Releases Gemini 3.5 Flash Frontier Model Focused On Agentic AI

It wouldn't be a Google I/O keynote without a new frontier AI model. At I/O 2026, Google finally revealed the Gemini 3.5 family to the world, starting with Gemini 3.5 Flash. As you'd expect, it's the latest and greatest model from Google, designed to work across a slew of Google products like the Gemini app, Android, and Google's Antigravity coding platform. Google says that the latest frontier model is designed especially for complex agentic workflows.

Google isn't the only company we're thinking about today, either. Lifehacker is launching a Big Guessing Game, featuring questions about what Apple might announce at its various events this year. The entire CNET Group is participating (CNET, Mashable, PCMag, ZDNET, and Lifehacker), and you can participate in as many of the games as you want.

Gemini 3.5 Flash rolls out across the globe

You won't need to wait for Gemini 3.5—at least not for Gemini 3.5 Flash, Google's frontier free model. It's rolling out today in the Gemini app, as well as Google Search's AI Mode. Developers can also access the 3.5 Flash model via Antigravity and Gemini API in Google AI Studio.

Google I/O's keynote and all the new AI features announced today are based on the 3.5 Flash model. Google was surprisingly quiet about the Gemini Pro model, but the company said it is testing it internally, and it will be rolled out next month.

According to Google, Gemini 3.5 Flash outperforms Gemini 3.1 Pro on coding and agentic benchmarks like Terminal-Bench 2.1 (76.2%), GDPval-AA (1656 Elo), and MCP Atlas (83.6%), and leads in multimodal understanding (84.2% on CharXiv Reasoning). When it comes to output tokens per second, the new model is four times faster than other frontier models.

Gemini 3.5 Flash is all about agentic AI and coding

Gemini 3.5 Flash's biggest enhancement comes in long-running and complex agentic workflows. Google has worked with partners and enterprise customers across fintech and data science teams to hone the agentic AI capabilities of the 3.5 Flash model. When it comes to agentic AI for regular consumers, Gemini 3.5 Flash will enable agentic browsing in Google Search's AI Mode and Gemini Spark, Google's new personal AI agent.

In AI Mode, Google is starting off with information agents. These agents work in the background, 24/7, collecting information for you. The agents can scour the web for news articles, social media posts, and tap into Google's real-time data for finance, sports, and shopping. For example, you can use search agents to notify you when your favorite athlete collaborates with your favorite sneaker brand for an exclusive drop. These features will be available for free across the globe.

Google is also testing Gemini Spark with select users, with a beta coming to Google AI Ultra subscribers next week (limited to the U.S.). Gemini Spark is a personal intelligence agent that works 24/7 in the background. It can even take actions on your behalf while under your direction. AI Mode will also include coding agents that can generate mini-apps and interactive dashboards for you on the fly, right in the Search interface.

Gemini Omni lets you generate videos using text and multimedia inputs

Google had a bit of a moment with Nano Banana 2.0: Image generation improved significantly, especially with text. Now, Gemini Omni can create media based on different kinds of input modalities. Google is starting off with video generation. In one prompt, you can combine images, audio, video, and text sources and use it to generate high-quality videos. Google will also let you edit videos in the same conversation.

Google will start rolling out Gemini Omni Flash to the Gemini app, YouTube Shorts, and Google Flow. Right now, it can only generate videos, but Google will add support for image and audio outputs down the line.

View the full article