Google I/O produced a flurry of AI announcements, as expected, and since it’s hard to keep up with everything they launched, here is a small recap.
Read on for the news in short on the new Gemini, the Omni video model, new subscription plans, the Gemini Spark agent and the revamped search box…
New Gemini model across Google
Gemini 3.5 Flash is now available to «billions of people worldwide,» being rolled out as the primary model for the Gemini app and AI Mode. It is also available in Antigravity, the API and in Android Studio — where it can build native Android apps.
The model is «a major leap forward in building more capable, intelligent agents,» Google says.
On benchmarks, it is largely on par with GPT-5.5, but nudges slightly ahead on agentic capabilities and financial tasks.
Gemini Omni takes on multimodal prompts
This is Google’s new video generation model, but it can do much more.
Omni is a truly multimodal model, and takes anything from any input — be it images, audio, video or text — and creates a video to your liking.
You can then edit the video through conversation.
It is rolling out now to Plus and Ultra subscribers and in Google Flow and YouTube Shorts.
New Ultra plan
Following Claude and ChatGPT, Google is announcing a subscription for the same tier in the $100 AI Ultra plan.
Targeting coders, knowledge workers and creators, the plan gives users a 5X higher usage limit on Gemini and Antigravity compared to the Pro plan.
At the same time, Google is lowering the price of the ultra Ultra plan from $250 to $200. This plan gets you 20x the usage limits of the Pro plan.
The Gemini Spark agent handles almost anything
Google then announced Gemini Spark — a new 24/7 agent to function as an «active partner.» It can connect to Google Workspace apps, including Gmail, where it can collect and summarize emails, and write up polished reports in Google Docs, for example.
It also connects to Canva, Openable and Instacart, with more apps promised within the coming weeks, when it also should be able to send texts and emails and operate a browser.
The agent is strictly opt-in, and you get to control which apps it connects to. It will also ask permission before spending money or sending emails.
More on: The Verge, TechCrunch, and Engadget.
Updating the search box
People have started asking Google a lot more complex questions with its AI search offerings, and to reflect that, Google is updating the search box.
The new bar looks the same at first glance, but now it can drop down to offer more lines of text, and you also get «AI-powered suggestions» for longer prompts.
The new bar also lets you input text, images, files, videos or even Chrome tabs.
The search results should be the same mixture of AI answers on top, and the ten blue links lower down, but complex queries will sometimes drop you straight into AI Mode.