Google used its I/O developers conference to lay out a broader push into “agentic” artificial intelligence, announcing new Gemini model rollouts and tools designed to handle tasks and produce media across Google’s products.
In the keynote, CEO Sundar Pichai told attendees near Google’s Mountain View, California headquarters that the company is “firmly in our agentic Gemini era,” describing the potential for agents while saying it remains early days to make them easy to use, “super secure and truly helpful.” He said Gemini’s app had 400 million monthly active users last year and that usership has now surpassed 900 million, more than doubling in a year. The conference setting also included a look at the company’s ongoing investment in AI, with Pichai pointing to growth in quarterly earnings and the stock’s gain after an investor call in late April in which Alphabet said capital expenditures may climb as high as $190 billion.
Google said the latest family of models, Gemini 3.5, is rolling out Tuesday to billions of users starting with Gemini 3.5 Flash. The company said Flash is focused on speed and characterized it as its strongest agentic and coding model yet, while also saying it is about four times faster than some competitors. Google said Gemini 3.5 Flash becomes the default model in the Gemini app and in “AI mode” on Google Search, and it is also working on a Gemini Pro 3.5 version it expects to launch next month.
Alongside the performance message, Google said its safety training and mitigations for Gemini 3.5 make the model “less likely to generate harmful content or to mistakenly refuse to answer safe queries.” The company also announced a new model, Gemini Omni, which it said would enable users to create high-quality video by entering a query using different inputs, including text, images, videos and audio, with follow-on editing “through a conversation with the model.” Google said it expects users will eventually be able to create images and audio with Omni, but it did not provide timing.
Google said Omni’s videos would look more realistic than videos made by other models because it understands physical effects including gravity, kinetic energy and fluid dynamics. It said Gemini Omni Flash, the first Omni model, launches Tuesday for Google AI Plus, Pro and Ultra subscribers through the Gemini app and Google Flow, and that starting this week it will be available at no cost on YouTube Shorts and the YouTube Create app. For provenance, Google said videos created with Omni will include Google’s imperceptible digital watermark, SynthID, and it said it is also adding content-credentials verification to the Gemini app. Google described that verification tool as determining whether content such as photos or videos was created by AI or captured with a phone camera and then edited with AI tools; it said that verification will appear in search in Chrome in the coming months. Google also said OpenAI, Kakao and Eleven Labs are adopting SynthID for their AI-generated content.
Google positioned the conference’s “agentic” theme around an assistant designed to act more autonomously than chatbots. Powered by Gemini 3.5, Gemini Spark, the company said, would be able to complete routine tasks such as sorting through meeting notes, emails and chats, then creating a document with the biggest takeaways and to-dos. Unlike some other agents, Google said Spark is based in the cloud so it continues working in the background even after a user shuts their laptop or locks their phone.
Because the shift toward proactive action has raised questions about how much autonomy users should grant, Google said Gemini Spark is designed to ask for permission before performing “high-stakes” tasks such as sending an email or making a purchase. Google said select testers would get access beginning Tuesday, and that it plans to roll out a beta mode to U.S.-based subscribers in its Google AI Ultra tier. The company also said that later this summer Spark will operate directly within Chrome.
Google also updated its devices roadmap at the conference, describing two kinds of smart glasses—audio glasses that provide spoken help and display glasses that offer information visually. Google said the audio version would come first, with arrival expected later this fall, and said users would be able to say “Hey Google” or tap the side of the frame to access Gemini for navigation, managing communications on a phone, real-time translations and other tasks.
In collaboration with Samsung and eyewear brands Gentle Monster and Warby Parker, Google showed initial designs for the glasses on Tuesday, including sunglasses from Gentle Monster and glasses from Warby Parker, with Google saying the designs will roll out as part of the brands’ full collections later this year. The company said it expects the audio glasses’ assistance to connect with other Gemini features once the devices reach consumers.
The conference also focused on expanding Gemini’s role in search and shopping. Google said “AI mode” introduced last year gives users more conversational answers before showing relevant links, and it said AI mode queries have more than doubled every quarter since its launch and recently surpassed 1 billion monthly users, according to Liz Reid, Google’s head of search. For the rollout, Google said the default model in search will now be Gemini 3.5 Flash, and it introduced an “intelligent search box” that it said is the biggest search box upgrade in 25 years. Google said the new box adapts to accommodate longer queries and helps users write out their questions with AI-powered suggestions instead of traditional autocomplete.
Google said the search box will also start rolling out Tuesday in all countries and languages where AI mode is already available, and that users can search using multiple modalities including text, images, video, files and even Chrome tabs as inputs. For shopping, Google announced a tool it called Universal Cart, which it described as an “intelligent shopping cart” that works across merchants and across services so users can add items to a cart while browsing Google Search, chatting with Gemini, watching YouTube or reading emails in Gmail. Google said Universal Cart will use Gemini models to look for deals and price drops, provide price history information and alert users when items are back in stock, and that it will be available on Search and the Gemini app this summer with YouTube and Gmail to follow.