AI Mode, Project Beam, Veo 3, Project Aura Glasses, Jules, and Everything Else Google Announced at I/O 2025

At Tuesday’s Google I/O 2025, the company introduced ambitious advancements in its quest to lead the development and commercialization of AI. Google showcased some of its most innovative features, including Gemini 2.5 Pro’s new reasoning mode called Deep Think, and AI Mode, its next step in redesigning search.

Gemini Ultra, Veo 3, and Image 4: A Subscription for Those Who Want It All

Gemini Ultra is Google’s most comprehensive AI subscription, priced at $249.99 a month.
The plan includes access to tools such as the Veo 3 video generator, the Flow editing app, and Gemini 2.5 Pro’s Deep Think mode, which isn’t available yet.
Subscribers also receive enhancements for NotebookLM and Whisk, up to 30 TB of cloud storage, YouTube Premium, and direct access to the Gemini chatbot via Chrome.
Many of the advanced features are powered by Project Mariner technology, which provides agents with “agentic” capabilities.

Deep Think: An AI That Takes Its Time to Respond Better

Deep Think is a new reasoning mode for Gemini 2.5 Pro that allows the AI model to consider multiple potential answers before selecting one.
This feature aims to enhance accuracy in handling complex tasks and advanced benchmarks. It’s currently available only to a select group of testers through the Gemini API.
Google is conducting security evaluations before making it publicly available.

AI Mode and Search Live: How Google Aims to Redesign Search

AI Mode is an experimental feature for Google Search that enables users to ask complex, multi-item queries through an AI-based interface.
This feature can handle sports and financial data. It also offers users the opportunity to try on clothes virtually.
Search Live will allow users to ask queries based on real-time detection by their mobile camera. It’s set to arrive over the summer.

The Way You Search Online Is Over. Google’s AI Mode Doesn’t Just Deliver Results—It Converses

Gemini in Chrome and SynthID Detection

Gemini has been integrated into Chrome as a browser assistant to help users better understand web page content and perform various tasks.
Gmail has introduced personalized smart replies and a new feature designed to help clean up the inbox.
Google has launched the SynthID Detector, a verification system that employs invisible watermarks to identify AI-generated content.

Beam: 3D Video Calls with Simultaneous Translation

Beam, previously known as Project Starline, transforms video calls into nearly face-to-face conversations using an array of six cameras and a light-field display.
It features millimeter-level head tracking and can deliver video at 60 frames per second.
It includes real-time translation in Google Meet, preserving the voice, tone, and expressions of the original speaker.

Jules: Google’s Assistant for Programming Without a Keyboard

Jules is Google’s new assisted programming agent, designed to compete with platforms such as Cursor, Windsurf, and Codex.
It can generate tests, update dependencies, write audio changelogs, and fix bugs while allowing users to continue working on other tasks.
Jules operates without the need for plugins or additional installations and is currently available in public beta.

Android Features Unveiled at the Event

Android debuts new tools for locating phones and lost objects, along with a new design language called Material 3 Expressive.
Google also showcased its new smart glasses, developed in collaboration with Xreal and based on Android XR.
The demo featured voice interaction with Gemini, simultaneous translation, and real-time information overlays.

This initiative is referred to as Project Aura, which aims to integrate Android into the XR space with a straightforward, practical approach.

Image | Google