Every thing you must know from Google I/O 2025

From the opening AI-influenced intro video set to “You Get What You Give” by New Radicals to CEO Sundar Pichai’s sign-off, Google I/O 2025 was filled with information and updates for the tech big and its merchandise. And once we say packed, we imply it, as this 12 months’s Google I/O clocked in at almost two hours lengthy. 

Throughout that point, Google shared some large wins for its AI merchandise, equivalent to Gemini topping numerous classes on the LMArena leaderboard. One other instance that Google appeared actually happy with was the truth that Gemini accomplished Pokémon Blue a number of weeks in the past.

However, we all know what you are actually right here for: Product updates and new product bulletins.

Except for a number of braggadocious moments, Google spent most of these 117 minutes speaking about what’s popping out subsequent. Google I/O mixes consumer-facing product bulletins with extra developer-oriented ones, from the newest Gmail updates to Google’s highly effective new chip, Ironwood, coming to Google Cloud clients later this 12 months.

We will break down what product updates and bulletins you must know from the total two-hour occasion, so you’ll be able to stroll away with all of the takeaways with out spending the identical time it takes to observe a serious movement image to study them.

Earlier than we dive in although, here is probably the most stunning information out of Google I/O: The subscription pricing that Google has for its Google AI Extremely plan. Whereas Google supplies a base subscription at $19.99 monthly, the Extremely plan is available in at a whopping $249.99 monthly for its complete suite of merchandise with the very best fee limits out there.

Google Search AI Mode

Google tucked away what is going to simply be its most seen characteristic manner too far again into the occasion, however we’ll floor it to the highest.

At Google I/O, Google introduced that the new AI Mode characteristic for Google Search is launching immediately to everybody in the US. Principally, it should permit customers to make use of Google’s search characteristic however with longer, extra complicated queries. Utilizing a “query fan-out technique,” AI Mode will be capable of break a search into a number of elements so as to course of every a part of the question, then pull all the data collectively to current to the person. Google says AI Mode “checks its work” too, however its unclear at the moment precisely what which means.

AI Mode is offered now. Later in the summertime, Google will launch Private Context in AI Mode, which is able to make recommendations primarily based on a person’s previous searches and different contextual details about the person from different Google merchandise like Gmail. 

As well as, different new options will quickly come to AI Mode, equivalent to Deep Search, which might dive deeper into queries by looking out by means of a number of web sites, and knowledge visualization options, which might take the search outcomes and current them in a visible graph when relevant.

In keeping with Google, its AI overviews in search are considered by 1.5 billion customers each month, so AI Mode clearly has the most important potential person base out of all of Google’s bulletins immediately.

AI Buying

Out of all of the bulletins on the occasion, these AI procuring options appeared to spark the most important response from Google I/O stay attendees.

Related to AI Mode, Google confirmed off its Buying Graph, which incorporates greater than 50 billion merchandise globally. Customers can simply describe the kind of product they’re searching for – say a particular sort of sofa, and Google will current choices that match that description.

Google additionally had a big presentation that confirmed its presenter add a photograph of themselves in order that AI might create a visible of what she’d appear to be in a gown. This digital try-on characteristic will probably be out there in Google Labs, and it is the IRL model of Cher’s Clueless closet.

The presenter was then ready to make use of an AI procuring agent to maintain tabs on the merchandise’s availability and observe its worth. When the worth dropped, the person obtained a notification of the pricing change.

Google stated customers will be capable of strive on completely different appears through AI in Google Labs beginning immediately.

Android XR

Google’s long-awaited post-Google Glass AR/VR plans have been lastly introduced at Google I/O. The corporate additionally unveiled various wearable merchandise using its AR/VR working system, Android XR.

One essential a part of the Android XR announcement is that Google appears to grasp the completely different use instances for an immersive headset and an on-the-go pair of smartglasses and have constructed Android XR to accommodate that.

Whereas Samsung has beforehand teased its Undertaking Moohan XR headset, Google I/O marked the primary time that Google revealed the product, which is being inbuilt partnership with the cell big and chipmaker Qualcomm. Google shared that the Undertaking Moohan headset needs to be out there later this 12 months.

Mashable Gentle Pace

Along with the XR headset, Google introduced Glasses with Android XR, smartglasses that incorporate a digital camera, audio system, and in-lens show that join with a person’s smartphone. Not like Google Glass, these sensible glasses will incorporate extra trendy appears because of partnerships with Light Monster and Warby Parker.

Google shared that builders will be capable of begin creating for Glasses beginning subsequent 12 months, so it is probably {that a} launch date for the smartglasses will comply with after that.

Gemini

Simply the star of Google I/O 2025 was the corporate’s AI mannequin, Gemini. Google introduced a brand new up to date Gemini 2.5 Professional, which it says is its strongest mannequin but. The corporate confirmed Gemini 2.5 Professional getting used to show sketches into full functions in a demo. Together with that, Google launched Gemini 2.5 Flash, which is a extra reasonably priced model of the highly effective Professional mannequin. The latter will probably be launched in early June with the previous popping out quickly after. Google additionally revealed Gemini 2.5 Professional Deep Suppose for complicated math and coding, which is able to solely be out there to “trusted testers” at first.

Talking of coding, Google shared its asynchronous coding agent Jules, which is presently in public beta. Builders will be capable of make the most of Jules so as to sort out codebase duties and modify information.

Builders may also have entry to a brand new Native Audio Output text-to-speech mannequin which might replicate the identical voice in numerous languages.

The Gemini app will quickly see a brand new Agent Mode, bringing customers an AI agent who can analysis and full duties primarily based on a person’s prompts.

Gemini may also be deeply built-in into Google merchandise like Workspace with Customized Sensible Replies. Gemini will use private context through paperwork, emails, and extra from throughout a person’s Google apps so as to match their tone, voice, and magnificence so as to generate computerized replies. Workspace customers will discover the characteristic out there in Gmail this summer season.

Different options introduced for Gemini embody Deep Analysis, which lets customers add their very own information to information the AI agent when asking questions, and Gemini in Chrome, an AI Assistant that solutions queries utilizing the context on the internet web page {that a} person is on. The latter characteristic is rolling out this week for Gemini subscribers within the U.S.

Google intends to carry Gemini to all of its gadgets, together with smartwatches, sensible automobiles, and sensible TVs.

Generative AI updates

Gemini’s AI assistant capabilities and language mannequin updates have been solely a small piece of Google’s broader AI puzzle. The corporate had a slew of generative AI bulletins to make too.

Google introduced Imagen 4, its newest picture era mannequin. In keeping with Google, Imagen 4 supplies richer particulars and higher visuals. As well as, Imagen 4 is seemingly significantly better at producing textual content and typography in its graphics. That is an space which AI fashions are notoriously dangerous at, so Imagen 4 seems to be a giant step ahead.

A brand new video era mannequin, Veo 3, was additionally unveiled with a video era device known as Circulate. Google claims Veo 3 has a stronger understanding of physics when producing scenes and may also create accompanying sound results, background noise, and dialogue. 

Each Veo 3 and Circulate can be found immediately alongside a brand new generative music mannequin known as Lyria 2.

Google I/O additionally noticed the debut of Gemini Canvas, which Google describes as a co-creation platform.

Undertaking Starline aka Google Beam

One other large announcement out of Google I/O: Undertaking Starline is not any extra.

Google’s immersive communication undertaking will now be often known as Google Beam, an AI-first communication platform.

As a part of Google Beam, Google introduced Google Meet translations, which mainly supplies real-time speech translation throughout conferences on the platform. AI will be capable of match a speaker’s voice and tone, so it seems like the interpretation is coming straight from them. Google Meet translations can be found in English and Spanish beginning immediately with extra language on the way in which within the coming weeks.

Google additionally had one other work-in-progress undertaking to tease below Google Beam: A 3-D conferencing platform that makes use of a number of cameras to seize a person from completely different angles so as to render the person on a 3-D light-field show.

Undertaking Astra

Whereas Undertaking Starline might have undergone a reputation change, it seems Undertaking Astra continues to be kicking it at Google, not less than for now.

Undertaking Astra is Google’s real-world common AI assistant and Google had loads to announce as a part of it.

Gemini Reside is a brand new AI assistant characteristic that may work together with a person’s environment through their cell machine’s digital camera and audio enter. Customers can ask Gemini Reside questions on what they’re capturing on digital camera and the AI assistant will be capable of reply queries primarily based on these visuals. In keeping with Google, Gemini Reside is rolling out immediately to Gemini customers.

It seems Google has plans to implement Undertaking Astra’s stay AI capabilities into Google Search’s AI mode as a Google Lens visible search enhancement.

Google additionally highlighted a few of its hopes for Gemini Reside, equivalent to with the ability to assist as an accessibility device for these with disabilities.

Undertaking Mariner

One other certainly one of Google’s AI tasks is an AI agent that may work together with the online so as to full duties for the person often known as Undertaking Mariner. 

Whereas Undertaking Mariner was beforehand introduced late final 12 months, Google had some updates equivalent to a multi-tasking characteristic which might permit an AI agent to work on as much as 10 completely different duties concurrently. One other new characteristic is Train and Repeat, which would offer the AI agent with the power to study from beforehand accomplished duties so as to full related ones with out the necessity for a similar detailed route sooner or later.

Google introduced plans to carry these agentic AI capabilities to Chrome, Google Search through AI Mode, and the Gemini app.