Google Has Released the Gemini 3.5 Flash Frontier Model, Which Focuses on Agent-Based AI.
A Google I/O presentation wouldn’t be complete without a new AI enabler. At I/O 2026, Google finally unveiled the Gemini 3.5 family, starting with Gemini 3.5 Flash. As expected, this is Google’s latest and greatest, designed to work with a variety of Google products, such as the Gemini app, Android, and the Google Antigravity programming platform. Google claims that this latest enabler is designed specifically for complex workflows involving agents.
Today, we’re not just thinking about Google. Lifehacker is launching “The Big Guessing Game,” which asks questions about what Apple might announce at its various events this year. The entire CNET team (CNET, Mashable, PCMag, ZDNET, and Lifehacker) is participating, and you can participate in as many games as you like.
The Gemini 3.5 Flash smartphone goes on sale worldwide.
You won’t have to wait for Gemini 3.5—at least not Gemini 3.5 Flash, Google’s free, innovative model. It’s already being implemented in the Gemini app and in Google Search’s AI mode. Developers can also access the 3.5 Flash model through Antigravity and the Gemini API in Google AI Studio.
The Google I/O presentation and all the new AI features announced today are based on the Flash 3.5 model. Google has been surprisingly quiet about the Gemini Pro model, but the company has stated that it is testing it internally and will be released next month.
According to Google, Gemini 3.5 Flash outperforms Gemini 3.1 Pro in programming and agent benchmarks such as Terminal-Bench 2.1 (76.2%), GDPval-AA (1656 Elo), and MCP Atlas (83.6%), and also leads in multimodal reasoning (84.2% in the CharXiv Reasoning test). In terms of tokens processed per second, the new model is four times faster than other leading models.
Gemini 3.5 Flash is all about self-organizing artificial intelligence and programming.
The most significant improvement in Gemini 3.5 Flash addresses the lengthy and complex workflows performed by agents. Google worked with partners and enterprise customers in fintech and data science to enhance the agent-based AI capabilities of 3.5 Flash. For user-facing agent-based AI, Gemini 3.5 Flash will enable agent-based searches in Google Search AI mode and Gemini Spark, Google’s new personal AI agent.
In AI mode, Google is starting to work with information agents. These agents work in the background 24/7, collecting information for you. They can search the web for news articles, social media posts, and use real-time Google data to find information on finance, sports, and shopping. For example, you can use search agents to get notified when your favorite athlete collaborates with your favorite sneaker brand on an exclusive collection. These features will be available for free worldwide.
Google is also testing Gemini Spark with select users, and the beta will be available to Google AI Ultra subscribers next week (US only). Gemini Spark is a personal intelligent agent that works 24/7 in the background. It can even perform actions on your behalf under your direction. In AI mode, human developers will also be available, allowing them to create mini-apps and interactive dashboards on the fly, right within the search interface.
Gemini Omni allows you to create videos using text and multimedia input.
Google has made significant progress with Nano Banana 2.0 : image generation, especially text generation, has improved significantly. Gemini Omni can now create media content based on various types of input. Google is starting with video generation. A single request can combine images, audio, video, and text, creating high-quality videos. Google will also allow video editing directly in the conversation.
Google will begin implementing Gemini Omni Flash in the Gemini app, YouTube Shorts, and Google Flow. Currently, it can only generate video, but Google will add support for image and audio output in the future.