Back
Technology

Google Unveils Gemini 3.5, AI Agents, Smart Glasses, and Unified Shopping Cart at I/O 2026

Generated on: Last updated: 17 sources
View source

Google I/O 2025: A New Era of AI Agents, Smart Glasses, and Universal Shopping

At its annual I/O developer conference, Google announced a sweeping suite of new artificial intelligence products and updates, including new AI models, an integrated AI agent, redesigned search capabilities, smart glasses, and a cross-platform shopping cart. Many features are initially rolling out to paying subscribers, with availability varying by region and plan.

New AI Models

Gemini 3.5 Family

Google launched the Gemini 3.5 family of AI models, beginning with Gemini 3.5 Flash. The model is now generally available as the default for the Gemini app, AI Mode in Google Search, the Gemini API, and the Antigravity coding assistant.

Google states the model generates output approximately four times faster than other frontier models and outperforms the previous Gemini 3.1 Pro model on coding, agentic, and multimodal benchmarks.

Benchmark results reported by Google for Gemini 3.5 Flash include:

  • 76.2% on Terminal-Bench 2.1
  • 1656 Elo on GDPval-AA
  • 83.6% on MCP Atlas
  • 84.2% on CharXiv Reasoning for multimodal understanding

The model was developed under Google's Frontier Safety Framework with new safety training methods designed to reduce harmful content generation and mistaken refusals. Gemini 3.5 Pro is scheduled for release next month.

Gemini Omni

Google announced the Gemini Omni model family, starting with Omni Flash. The model accepts text, image, audio, and video input and outputs editable video. Google states the model incorporates knowledge of gravity, kinetic energy, and fluid dynamics.

Omni Flash is available to Google AI Plus, Pro, and Ultra subscribers in the Gemini app and Google Flow. It will be available for free to all users on YouTube Shorts and the YouTube Create App.

AI Agents

Gemini Spark: Your Persistent Cloud Agent

Gemini Spark is a persistent AI agent that operates on dedicated servers and is accessible through the Gemini app. It integrates with Google Workspace applications including Gmail, Docs, Slides, and Calendar, as well as third-party services via MCP connections (including Canva, OpenTable, and Instacart). The agent can complete tasks such as sorting meeting notes, creating documents, and managing schedules.

The agent operates in the cloud while the user's device is idle, including when locked. It requires user permission before executing high-stakes actions such as sending emails or making purchases.

Beta access is available next week for US Google AI Ultra subscribers (priced at $100/month). Google plans to bring agentic features to free users at a later date.

Information Agents & Android Halo

A separate feature called "information agents" will monitor the web for updates on specified topics and return summaries or action plans. This feature is scheduled to launch this summer for US Google AI Pro and Ultra subscribers.

Android Halo is a notification system that provides at-a-glance visibility of agent activity on the phone screen. It will be available later this year for Gemini Spark and other supported agents.

Google Search Updates

AI Mode and Search Redesign

AI Mode in Google Search now defaults to Gemini 3.5 Flash. The search box has been redesigned to expand as users type, accommodating longer, more specific queries. Users can input text, images, files, videos, and Chrome tabs as references. Web links are provided below AI-generated output.

Google reports AI Mode queries have more than doubled every quarter since its launch and now serve over 1 billion monthly users. Search queries reached an all-time high according to Alphabet CEO Sundar Pichai and vice president Liz Reid.

Users can still access traditional link-based Search through a "Web" tab. Later in 2026, Google states the search box will generate interactive visualizations, tables, and graphics. Paying subscribers will be able to build mini-apps for recurring tasks within Search.

Personal Intelligence

Users can connect AI Mode to Personal Intelligence, allowing the AI to access data from Gmail, Google Photos, and other Google services.

Gemini App Updates

Daily Brief

A new feature called Daily Brief creates a morning digest using information from connected apps such as Gmail, Calendar, and Tasks. It prioritizes tasks and suggests next steps. The feature is rolling out to Google AI Plus, Pro, and Ultra subscribers in the US.

Neural Expressive Redesign

The Gemini app has been redesigned with a new visual language called "Neural Expressive," featuring fluid animations, vibrant colors, haptic feedback, and new typography. Gemini Live voice mode is now integrated directly into the app, allowing users to switch between typed prompts and voice conversations. The redesign is rolling out globally across web, Android, and iOS.

Gemini for macOS

A desktop app for macOS is available for download. Spark and new voice features for the desktop app are scheduled to roll out later this summer.

Commerce: Universal Cart

Google introduced Universal Cart, an AI-powered shopping cart that aggregates products across Google Search, the Gemini app, YouTube, and Gmail. The tool can track price drops, show price history, alert users when items return to stock, and flag product compatibility issues.

It is built on Google Wallet and accounts for payment perks, loyalty information, and merchant offers. Users can check out with Google Pay or transfer items to a merchant's site.

According to Google, shopping activity across its services occurs more than one billion times per day, and the Shopping Graph includes more than 60 billion product listings. Universal Cart will roll out in the US this summer, beginning with Search and the Gemini app, followed by YouTube and Gmail.

Google will also bring its Agent Payments Protocol to its products starting with Gemini Spark, allowing users to set spending limits and purchase criteria before an agent completes a transaction.

Smart Glasses: Android XR and Project Aura

Google announced two types of smart glasses developed in partnership with Samsung, Qualcomm, Gentle Monster, and Warby Parker.

Audio-Only Glasses

  • Includes speakers, microphones, and cameras but no display screen
  • Users can summon Gemini by tapping the frames or using voice commands
  • Features include directions, photo and video capture, music playback, calling, message transcription, and live translation
  • Launch this fall through Gentle Monster (sunglasses) and Warby Parker (prescription glasses)
  • Can be paired with Android phones and iPhone

Display-Equipped Glasses (Project Aura)

  • Developed with Xreal featuring an in-lens display that can overlay up to five apps on the real world
  • Tethered to a redesigned processor box with a fingerprint sensor
  • Can serve as an external monitor for other devices

Content Credentials and SynthID

Google's SynthID digital watermark has been embedded in over 100 billion images and videos, plus over 500 million hours of audio. The company stated that OpenAI, Kakao, and Eleven Labs are adopting SynthID.

Google is rolling out a Content Credentials verification feature using the C2PA standard, allowing users in Gemini, Search, and Chrome to inquire about the provenance of media, including whether it was AI-generated or edited. Users can right-click an image or use Circle to Search to check this information.

Workspace and Creative Tools

Docs Live, Gmail Live, and Keep

  • Docs Live allows users to draft documents by speaking to Gemini, with the draft updating in real time. Available on Android and iOS this summer for AI Pro and Ultra subscribers.
  • Gmail Live is a conversational email search and draft feature rolling out to AI Pro and Ultra subscribers in the US on Android and iOS this summer.
  • Google Keep receives a similar voice-driven note organization feature.

Google Pics & Google Beam

Google Pics is an AI image creation and editing tool for Workspace, powered by Nano Banana 2 and Gemini models. It will be initially available for paid Workspace users.

Google demonstrated "Sophie," a lifelike AI agent for Beam calls that responds to questions, reads documents, and provides recommendations. Group calls using Beam were also shown.

Developer and Coding Tools

Antigravity 2.0

Google introduced an updated version of Antigravity, its coding assistant. Antigravity 2.0 is a standalone desktop application for managing multiple agents. Google also announced Antigravity CLI for terminal users and an Antigravity SDK for custom agent behavior.

AI Studio and Android

Google announced features for building native Android apps using AI, including an embedded emulator in AI Studio and the ability to publish apps to the Play Store. Apps can be exported to Android Studio or GitHub.

Managed Agents

In the Gemini API, Google introduced Managed Agents that provide a remote Linux environment for agents to browse the web, execute code, and manage files.

YouTube

Ask YouTube

A new feature called "Ask YouTube" handles complex search queries and lists relevant videos. It is available for YouTube Premium subscribers in the US.

Gemini Omni on YouTube

Gemini Omni will be available in YouTube Shorts Remix and the YouTube Create app.

Pricing Changes

Google reduced the price of its AI Ultra subscription from $250/month to $200/month. A $100/month version of Ultra aimed at developers and work-related users was also announced, including higher usage limits and 20TB of cloud storage. Google AI Plus subscription is available at $20/month.

Large enterprises could save more than $1 billion annually by switching to Google's models, according to CEO Sundar Pichai, who stated they deliver "comparable performance to rival models at roughly one-third of the cost."

Platform Reach

The Gemini app has grown from approximately 400 million to over 900 million monthly active users over the past year. The app is available in more than 230 countries and over 70 languages.

Expert and Leadership Commentary

"When we look back at this time, I think we will realize that we were standing in the foothills of the singularity. It will be a profound moment for humanity."
— Demis Hassabis, Head of Google DeepMind

"Making agents easy to use, secure, and helpful is still in early stages."
— Sundar Pichai, Alphabet CEO

  • Clarence Lee, Visiting Lecturer at Cornell SC Johnson College of Business, described AI agents as tools for delegating tasks such as composing emails or booking hotels.
  • Karan Girotra, Professor at Cornell University, noted that Gemini Spark's integration with Google apps gives it access to user context, which could improve performance.