12 Google I/O 2026 announcements you need to know about

O Google I/O 2026 This marked Google's clearest bet yet on the so-called "agent era." See the main announcements and what changes for users, developers, and those following the AI ​​race:

The official scale behind Google's AI.

During the presentation, Sundar Pichai, Google's CEO reinforced that Google's strategy is... "full-stack"In other words, it's complete: its own chips, infrastructure, Gemini models, and products used by billions of people. The company says it has... 13 products on more than 1 billion users, with five exceeding 3 billion. He further explains that the AI-generated Overview in Google search already totals 2,5 billion monthly active users.

Another important metric for measuring adoption is the volume of tokens: it was 9,7 trillion per month two years ago, then dropped to around 480 trillion in the last I/O, and now it's reaching... 3,2 quadrillion tokens per month on Google platforms. The company also mentions 8,5 million developers creating with their models every month and processing APIs around 19 billion tokens per minute.

1. Docs Live

O Google Docs (Google Docs, in English) won the role Docs LiveThis makes it easy to create and edit documents instantly using voice commands. In the video above, you can see how it understands the user and can translate even complex conversations into a document. Docs Live It will be released to subscribers in mid-2026, and the advanced voice features will also be coming to Gmail and Keep soon.

2. Gemini 3.5 Flash

According to Google, the newly announced Gemini 3.5 Flash It outperforms the Gemini 3.1 Pro in virtually all benchmarks cited by the company and delivers responses up to... four times fasterin tokens per second, which other high-end models... Pichai also highlighted the cost: Flash would offer advanced features for less than half the price of comparable models, allowing for significant savings in enterprise AI workloads.

O Gemini 3.5 Flash It will be available from today (19) and will become the standard AI model in the Gemini app and in Google Search AI Mode. The promise is faster responses, better task execution by agents, more programming capacity and generation of richer web interfaces. It has also received security improvements to reduce harmful responses and decrease false blocks of legitimate questions. Learn more about Gemini 3.5 Flash here..

3. New Gemini app

In addition to the new model, the app of Gemini: received a visual redesign that Google called Neural ExpressiveThe interface features new animations, more color, new typography, and haptic feedback. The update began rolling out on May 19th to the Android and iOS apps, as well as the web version.

4. Gemini Omni

In the official announcement, Google summarizes the Gemini Omni as a model capable of generating outputs in any modality from any type of input (text, photos, video, or audio). The debut begins with video featuring the Gemini Omni Flash, available on the Gemini app, Google Flow, and YouTube Shorts, with APIs for developers and businesses coming in the following weeks.

The difference compared to Veo is precisely that it accepts multiple input types, not just text. Google says that, in the future, Omni should be able to achieve this. "create anything from any input"Here's how it works:

5. TPU 8t and TPU 8i

The presentation also provided an important explanation of the technical basis for these features. Google states that its 8th generation of TPUs has been divided into two chips: the TPU 8t, optimized for large-scale training, and the TPU 8i, focused on inference. The promise is to train larger models in weeks, not months, and deliver up to Twice the performance per watt.

6. Gemini Spark

O Gemini Spark It's Google's bet on always-on personal agents. It uses Gemini 3.5 and resources of Google Antigravity To execute long-running tasks in the background, on dedicated virtual machines in Google Cloud. The idea is that the user can delegate actions on command — such as organizing emails, preparing documents, monitoring billing, or tracking tasks — without leaving their computer open.

The tool starts out integrated with Google products, but should also communicate with third-party services via MCPOn Android, the progress of agents like Spark will appear in the interface. Android HaloBy mid-2026, the feature should also arrive in Chrome as an agentic browsing experience. The Beta version will arrive soon, first for plan subscribers. Google AI Ultra USA.

7. Google AI Studio creates Android apps via prompt.

Google also introduced a feature of vibe coding to create native Android apps within Google AI StudioThe tool includes a built-in Android emulator for viewing and editing previews, allows testing the app on a connected mobile phone, and exporting the project to Android Studio, GitHub, or a ZIP file.

According to the company, in the future it will be possible to publish apps created via prompt to friends and family, without necessarily making them publicly available on the Play Store. Integrations with Firebase also planned.

8. Smart glasses with Android XR

Google has spoken out again about Android XR and showed new things for smart glassesThe highlight was... Project Aura, glasses with a screen designed in partnership with Xreal, and for two models of smart “audio glasses” created with Warby parker e Gentle Monster, in collaboration with Samsung:

Unlike glasses with screens, these models are described as audio glassesIn other words, smart glasses that communicate with you through audio from side speakers and built-in microphones. They will support live translation, Gemini navigation assistance, notification summaries, and camera functionality. Learn more about them. in this matter.

9. Universal Cart

Regarding the news from Google searches for purchases, the company Universal Cart, a universal shopping cart that allows you to add products from YouTube, Search, Gemini, and Gmail. The idea is to bring together items from different stores, integrating with them and even large marketplaces, and finalize the purchase in a more integrated experience.

Search engines will now have "purchase agents" that help users save product searches, notify them about discounts, and even make purchases autonomously.

Another advantage is the analysis of purchased items, to warn about problems or incompatibilities. For example, when buying components for a gaming PC, the AI ​​can alert you to incompatibility between the selected products, preventing the wrong purchase. In addition, it can also alert you to benefits from cards and loyalty programs registered in Google Wallet. Universal Cart It will first arrive on Google Search and Gemini in the Northern Hemisphere summer, with integration into YouTube and Gmail in the following months.

Learn more about Universal Cart in this complete article..

10. Gmail Live

O Gmail Live It brings a similar experience to Gemini Live for searching emails. Instead of typing terms and searching through conversations, the user can tap an icon in the search bar and ask questions by voice. The interface returns relevant information extracted from the inbox, such as a booking code or confirmation details.

Similar features will also be coming to Google Docs and Keep, with the ability to pull data from Drive and Gmail.

11. Pics: the image editor with prompts

O Google workspaceGoogle's suite of solutions for businesses will gain the PicsThis is an AI-powered image editing app. The idea is to click on a part of the image and leave a comment describing the desired change, instead of writing a long prompt explaining everything. The feature uses Gemini and Nano Banana 2, and should be incorporated into other Workspace apps in the future.

12. New Google Search

A Google search It has received its biggest overhaul in over 20 years. In practice, it is moving towards becoming an environment that tracks tasks using agents. Among the new features are information agents that track topics and summarize updates from blogs, news, and social networks; a generative interface capable of creating tables, graphs, and simulations; and mini-apps for recurring tasks, such as a personalized dashboard for planning events.

Learn more about New features of Google Search in this article..

Watch the video: Google I/O '26 Keynote

See also other features

Android Show 2026: See everything Google announced.

Two smart glasses from Samsung and Google will be demonstrated at I/O 2026.

Sources: Google Brazil/Sundar Pichai at I/O 2026, Google I/O 2026, Google Android XR, Google AI Ultra, The Verge, The Verge/Gemini, The Verge/Search e The Verge/Beam.