Gemini 3.5 Flash and Pro Models

Google launched the next generation of its AI models at I/O 2026, starting with Gemini 3.5 Flash, which is now the default model for the Gemini app and AI Mode in Search. The company claims it is significantly faster and better at handling agentic tasks, coding, and generating richer web UIs and graphics. Improved guardrails reduce harmful content generation and false positives on safe queries. The Gemini 3.5 Pro model is expected to follow next month, offering even more advanced capabilities for complex tasks.

Gemini App Redesign

Alongside the new model, Google unveiled a visual redesign for the Gemini app called “neural expressive.” It features new animations, pops of color, a fresh font, and haptic feedback. The update rolls out starting May 19th on the web, Android, and iOS. This redesign aims to make interactions with the AI feel more natural and engaging, aligning with Google's broader push to integrate AI more deeply into everyday tools.

Gemini Omni Models

Google introduced an entirely new family of AI models called Gemini Omni. The first model, Omni Flash, can generate video clips from prompts combining text, photos, video, and audio—a step beyond text-to-video models like Veo. It is rolling out in the Gemini app, Google Flow, and YouTube Shorts. Future versions promise to “create anything from any input,” potentially revolutionizing content creation across media types.

Gemini Spark – Always-On AI Agent

Described as Google's version of OpenClaw, Gemini Spark is an always-on AI agent powered by Gemini 3.5 Flash. It runs 24/7 on Google Cloud virtual machines and can connect to Google Workspace apps (Docs, Gmail, Sheets, Slides) and third-party apps like Canva and Instacart. Spark can write emails, create study guides, monitor for hidden credit card fees, and more. It will soon gain access to local files via the Gemini app on macOS, expanding its utility for personal and professional tasks.

AI Studio – Vibe-Coding Android Apps

Google AI Studio now allows users to build full native Android apps using natural language prompts. The feature includes an embedded Android emulator for previewing apps, and users can install apps directly on their phones from AI Studio. Apps can be exported to Android Studio, GitHub, or saved as ZIP files. Google plans to allow publishing apps exclusively for friends and family, as well as future Firebase integrations, making it easier for non-developers to create and share Android applications.

Project Aura Smart Glasses

Google collaborated with Xreal on an updated version of its Project Aura smart glasses. The external compute puck has been redesigned with a fingerprint sensor and a lanyard for portability. New features include widgets for display glasses, Gemini integrations with Google Calendar and Google Keep, and improved Gemini performance. The Verge's Victoria Song got an early look, noting the refinements suggest Google is serious about bringing augmented reality to consumers.

Android XR Glasses from Warby Parker and Gentle Monster

Two new pairs of Android XR smart glasses will launch this fall from Warby Parker and Gentle Monster. Unlike the display-equipped Project Aura, these are audio-only, similar to Ray-Ban Meta glasses. They support live translation, navigation assistance with Gemini, and notification summaries. The partnerships, announced last year, now have concrete designs and feature sets, indicating Google's commitment to the XR ecosystem beyond its own hardware.

Universal Cart – One Checkout for Everything

Google introduced a “Universal Cart” that lets users add products from YouTube, Search, Gemini, and Gmail and check out from multiple merchants (Nike, Target, Walmart, Ulta, Sephora, Wayfair, Shopify) simultaneously. The intelligent cart can flag potential issues, such as incompatible parts for a gaming PC, and interpret loyalty info from Google Wallet to help users find savings. It launches this summer in Search and Gemini, with YouTube and Gmail support coming later.

Gmail Live – Voice Search Your Inbox

Gmail is getting a voice-driven search feature called Gmail Live. Clicking an icon in the search bar lets users ask questions by voice, and instead of returning a list of emails, the interface extracts and delivers pertinent information—like a hotel confirmation code. Similar features will come to Google Docs and Keep, pulling data from Drive and Gmail. This aims to make information retrieval faster and more intuitive, especially on mobile devices.

Google Workspace App 'Pics' – AI Image Editing

Google Workspace gains a new app called Pics, powered by Nano Banana 2 and Gemini. Users can make iterative AI-generated updates to images by clicking on specific areas and leaving comments describing edits, eliminating the need for full prompts. Google plans to integrate Pics' capabilities into other Workspace apps, streamlining image editing across documents, slides, and more.

Search Upgrades: Agents, Generative UI, Mini Apps

Google revamped the search box to accommodate longer queries and include AI-generated suggestions. Users can now input text, images, files, videos, and Chrome tabs. “Information agents” will provide summarized updates on specific topics from blogs, news, social media, etc., launching this summer for AI Pro and Ultra subscribers. Generative UI can create visuals like simulations and interactive tables, while “mini apps” can be generated for repeated tasks, such as event planning dashboards.

AI Ultra Subscription Price Cut

Google reduced the price of its premium AI Ultra subscription from $249.99 per month to a starting tier of $100 per month, with a $200 tier that includes access to Project Genie. This move mirrors OpenAI's pricing strategy and makes advanced AI capabilities more accessible to a wider range of users, including higher usage limits and early access to new tools.

AI Detection Tools Expanded to Chrome and Search

To help users identify AI-generated or altered images, Google expanded its SynthID watermarking and C2PA Content Credentials to Chrome and Search. Uploading or selecting images in Search (including Lens and Circle to Search) can reveal provenance details. Later, Chrome will allow circling questionable images on websites to check their authenticity. This is part of Google's broader effort to combat misinformation and increase transparency in the age of generative AI.

Google Beam AI Agents

After renaming Project Starline to Google Beam last year, Google is experimenting with lifelike AI agents that can interact during calls. The Verge's Sean Hollister tried an early demo of Sophie, a Beam video agent that responds to questions, reads documents held up to the camera, and looks up information like restaurant recommendations. Google also demonstrated group calls using Beam, which will integrate with Google Meet and Zoom. These advances point toward a future where AI avatars can participate in video calls as naturally as humans.

Source: The Verge News