Sundar Pichai, chief govt officer of Alphabet Inc., in the course of the Google I/O Builders Convention in Mountain View, California, US, on Tuesday, Might 19, 2026.
David Paul Morris | Bloomberg | Getty Pictures
Google is rolling out its newest model of Gemini and a brand new synthetic intelligence mannequin designed to simulate the bodily world, because the search large races to maintain tempo in mannequin improvement whereas additionally offering extra agentic providers to its large person base.
The corporate made the bulletins at its annual Google I/O developer convention on Tuesday, gaining an viewers for brand spanking new product debuts at a time when the market has been centered on the hovering valuations of OpenAI and Anthropic, that are each gearing up for IPOs as quickly as this yr.
The centerpiece of Google’s AI technique is Gemini, its household of fashions and instruments. The corporate is showcasing Gemini 3.5 Flash, a lighter-weight addition to its suite that gives cutting-edge capabilities at half, or in some circumstances near one-third, the worth of comparable frontier fashions, based on CEO Sundar Pichai.
In a information briefing with reporters forward of Tuesday’s occasion, Pichai mentioned Gemini 3.5 Flash is “remarkably quick.” The corporate mentioned on Tuesday that 3.5 Flash will now be the default mannequin for the Gemini app and AI mode in search globally.
“You now not must commerce high quality for latency,” Google mentioned in a weblog put up. The corporate mentioned that it is strengthened the cybersecurity defenses for Gemini 3.5 Flash, so it is “much less prone to generate dangerous content material and mistakenly refuse to reply protected queries.”
Google mentioned Gemini 3.5 Professional, its heavier-weight model, is getting used internally, however will not be prepared for wider distribution till subsequent month.
On the agentic AI entrance, Google introduced Gemini Spark, a brand new common function AI agent within the Gemini app that may purpose throughout info in linked apps. Google mentioned it desires to assist customers navigate their digital lives by taking “motion in your behalf whereas below your route.” Gemini Spark is in beta and will likely be accessible first to trusted testers and Google AI Extremely subscribers, beginning subsequent week.
With extra web customers gravitating to chatbots, Google is attempting to persuade conventional search customers that it may be trusted to assist them with duties involving minimal enter. Following the corporate’s skyrocketing capital spending, Wall Road is in search of Google to indicate it could actually create deeper integrations throughout its merchandise, and brokers could possibly be a means to try this.
Expectations for AI corporations proceed to develop, notably in gentle of Anthropic’s not too long ago launched Mythos mannequin, which was mentioned to be so highly effective that it is discovered hundreds of beforehand unknown vulnerabilities on the planet’s software program infrastructure.
Google’s AI portfolio now consists of Omni, a world mannequin designed to simulate bodily environments, predicting what occurs subsequent primarily based on a person’s actions. World fashions are sometimes utilized in robotics and gaming and have been closely researched by DeepMind by way of the years.
Omni will work in Flash, the Gemini App, Google Stream and YouTube shorts, supporting picture and audio, the corporate mentioned, including in a separate weblog put up that customers could make Omni edit movies and create extra reasonable imagery.
“Take a video you shot and simply ask Omni to alter what’s taking place,” the put up says. The AI can “edit the motion, add in new characters or objects.”
WATCH: What to anticipate from Google I/O
