Everything Google announced at I/O 2025: Search’s AI mode, 3D video calls, Android smart glasses, and more |

Google’s annual developer convention, the I/O 2025, arrived with its typical flurry of bulletins, demos, and forward-looking guarantees. This yr’s I/O, held at the Shoreline Amphitheatre in Mountain View, continued the corporate’s all-in strategy on synthetic intelligence with vital updates throughout its product lineup. From an eye-popping $250/month Ultra subscription to Search reimagined with AI, Google delivered a smorgasbord of bulletins that ranged from genuinely spectacular to barely regarding to your pockets. Let’s do a run down of the whole lot that Google announced at the I/O 2025.

Gemini 2.5 fashions get even smarter

Google’s Gemini, the AI model powering most of the its providers, acquired main upgrades throughout its lineup. The new Gemini 2.5 Pro now options an “enhanced reasoning mode” referred to as Deep Think, which permits the mannequin to think about a number of hypotheses earlier than responding – notably helpful for advanced math and coding issues. According to Google, this places it at the highest of the LMArena leaderboard in all classes, with Elo scores up more than 300 factors because the first-generation Gemini Pro.Meanwhile, Gemini 2.5 Flash – the more environment friendly, cost-effective model – has been improved throughout reasoning, multimodality, code, and lengthy context capabilities. Google claims it is now second solely to 2.5 Pro on benchmark exams, whereas requiring 20-30% fewer tokens for responses. For builders on the lookout for highly effective, budget-friendly AI instruments, that is the equal of getting premium efficiency at a mid-range value level.

Project Starline will get actual and a brand new title, the Google Beam

Remember Project Starline, Google’s futuristic 3D video conferencing sales space that promised to make it appear to be your dialog companion was sitting proper throughout from you? It’s now not only a analysis challenge – it is turning into an actual product referred to as Google Beam.The know-how makes use of a six-camera array and AI to merge video streams into a practical 3D expertise seen on a light-weight subject show. Google claims “near perfect head tracking, down to the millimeter” at 60 frames per second, all occurring in real-time. HP would be the first to launch Beam units for enterprise prospects later this yr, with corporations like Deloitte, Duolingo, and Salesforce already signed up.I’m not saying it’ll exchange Zoom in a single day, however it would possibly really make these infinite digital conferences barely much less soul-crushing.

AI Ultra may very well be the subscription service nobody requested for however they could need it

In the “things nobody saw coming” class, Google announced a brand new premium subscription tier referred to as AI Ultra, priced at a jaw-dropping $250 monthly. Yes, you learn that accurately, 2 hundred and fifty {dollars}. Monthly.For this Netflix-annual-subscription-times-twenty price ticket, subscribers get early entry to Google’s newest AI instruments and limitless use of options like Deep Research. The package deal consists of 30TB of storage throughout Google Photos, Drive, and Gmail, plus YouTube Premium and entry to experimental options like Project Mariner. New subscribers can get 50% off for the primary three months, which nonetheless lands at $125 month-to-month.Google clearly believes some persons are keen to pay premium costs for cutting-edge AI instruments. The remainder of us will simply have to attend for these options to finally trickle all the way down to more reasonably priced tiers.

Google Search will get a whole makeover with AI mode

After what Google referred to as “one of the most successful launches in Search in the past decade” with AI Overviews, the corporate goes all-in on reworking how we search the net with AI Mode. This separate tab in Google Search lets customers ask longer, more advanced questions and comply with up with extra queries – primarily turning Google Search right into a conversational AI expertise.Early testers have been asking queries which are two to 3 instances longer than conventional searches, suggesting customers are already adapting to this new paradigm. AI Mode is rolling out to everybody within the US beginning this week, powered by Gemini 2.5.Google can also be including options like Deep Search, which expands background queries from tens to tons of to create complete search responses, and an AI purchasing expertise that helps customers discover objects and even nearly “try on” clothes by importing a photograph of themselves.

Google brings smart glasses out of shade with Android XR

Following final week’s Android Show, the place Google previewed some Android 16 options, I/O offered more particulars on Google’s blended actuality plans. The firm unveiled its Android XR platform for augmented, blended, and digital actuality units.The most attention-grabbing growth is Project Aura, a prototype of Android XR-powered smart glasses developed with Xreal. These glasses will characteristic Gemini integration and a big subject of view, together with built-in cameras and microphones. Google can also be partnering with widespread eyewear manufacturers together with Samsung, Gentle Monster, and Warby Parker to create more trendy choices.Android XR will help options like reside translation, directional navigation through a mini Google Maps show, and the flexibility to view immersive 360-degree movies. After Google Glass and earlier AR makes an attempt, this seems like the corporate’s most critical push but into wearable computing.

Gmail will get AI-powered Smart Replies that really sound such as you

Google’s AI is coming to your inbox, however in a useful approach. Gmail will quickly characteristic personalised smart replies that analyze your writing model and previous emails to recommend responses that sound authentically such as you. The system considers your typical greetings, tone, and even favourite phrase decisions to generate more related replies.With your permission, Gemini can pull info from each your Gmail and Google Drive to craft these responses. For occasion, if a good friend emails asking a couple of highway journey you have taken, Gemini might reference your previous itineraries saved in Docs to recommend an in depth response.The characteristic will launch by way of Google Labs in July for English customers on internet, iOS, and Android platforms. It’s both extremely handy or barely unnerving, relying on how you are feeling about AI mimicking your writing model.

Google Meet provides real-time AI translation

Video calls with worldwide colleagues or associates are about to get a lot simpler. Google Meet is including an AI-powered real-time translation characteristic that converts your speech into your dialog companion’s most popular language practically instantaneously. The system even makes an attempt to match your tone and cadence whereas translating.The characteristic initially helps translation between English and Spanish, with more languages deliberate for the close to future. It’s rolling out in beta to Google AI Pro and Ultra subscribers first. This may very well be a game-changer for world companies and multilingual households alike.

Google’s AI Agent is turning into more highly effective

Google’s experimental AI agent, Project Mariner, acquired vital upgrades. The system can now deal with as much as 10 completely different duties concurrently, from reserving flights to researching matters to evaluating purchasing choices.The newest model additionally introduces a “teach and repeat” perform, the place you may show a process as soon as, and it learns the way to carry out related duties sooner or later. Google is bringing Project Mariner’s capabilities to builders through the Gemini API, with trusted testers like Automation Anywhere and UiPath already constructing with it.

Google’s imaginative and prescient for a common AI assistant takes form with Project Astra

Remember when AI assistants have been simply glorified climate forecasters and timer-setters? Project Astra, Google’s bold common AI assistant prototype, is aiming to make these days really feel like historical historical past. At I/O 2025, Google showcased Astra’s enhanced capabilities, demonstrating the way it can now proactively use your telephone’s digicam to “see” the world round you and take motion with out specific instructions. Google is positioning Astra because the fruits of its AI assistant work, able to the whole lot from diving into your emails to seek out bike specs to researching restore info and even calling native retailers in your behalf. While some would possibly really feel queasy about giving an AI this a lot entry to their digital lives, the productiveness beneficial properties may very well be substantial, for those who’re comfy with an AI assistant which may sometimes pipe up with unsolicited recommendation when you’re making an attempt to repair your bike chain in peace. The Astra options are already showing in Gemini Live, which is now out there to all Android customers and rolling out to iOS beginning at the moment.

Google’s Veo and Imagen will get actual(er)

If you thought artificially generated movies look comedian, it is best to have watched Google’s keynote, because it’s new fashions have been previewed. Veo 3, the newest model of Google’s AI video generator, now helps audio era – including ambient sounds, music, and even dialogue to AI-generated movies. Previous variations might solely create silent clips, which wasn’t notably helpful in a world the place TikTok and YouTube dominate.Imagen 4, Google’s text-to-image mannequin, has been improved to generate more photorealistic pictures with higher dealing with of high-quality particulars like materials, water droplets, and animal fur. It can now export pictures in varied codecs and resolutions as much as 2K.Google can also be introducing Flow, an AI filmmaking app that builds on these applied sciences. Flow lets customers create eight-second AI-generated video clips primarily based on textual content prompts or pictures, with scene-builder instruments to sew clips collectively for longer movies. Think of it as having a mini movie manufacturing studio in your pocket – albeit one which typically has weird interpretations of your artistic imaginative and prescient.

Stitch is an AI-generated UI designer

For builders and designers, Google unveiled Stitch, an AI-powered software that generates person interfaces primarily based on textual content descriptions. Users can present wireframes, tough sketches, or screenshots of different designs to information Stitch’s output, making it simpler to shortly prototype app and web site designs.While it has more restricted capabilities than some competing AI design instruments, Stitch gives HTML and CSS markup for the designs it generates, serving to bridge the hole between idea and implementation. The experiment is at present out there by way of Google Labs.