Google I/O 2025 in a nutshell

Google hosted its annual I/O developer conference at Mountain View’s Shoreline Amphitheatre, unveiling a slew of updates across its portfolio, including Android, Chrome, Google Search, YouTube, and its AI-powered chatbot, Gemini, over two days starting Tuesday.

Gemini Ultra and 2.5 Pro Deep Think

Google announced Gemini Ultra, a premium subscription ($249.99/month, U.S.-only for now) offering unparalleled access to its AI apps and services. Subscribers gain access to Veo 3, a video generator capable of creating sound effects, dialogue, and high-quality footage, surpassing its Veo 2 predecessor. Additionally, Gemini Ultra includes Flow, a video editing app; Gemini 2.5 Pro Deep Think mode, an enhanced reasoning capability (currently in trusted tester phase awaiting safety evaluations); higher limits on NotebookLM and Whisk; agentic tools via Project Mariner; YouTube Premium; and 30TB of storage across Google Drive, Photos, and Gmail.

The Gemini app now has over 400 million monthly active users. Updates include the rollout of Gemini Live’s camera and screen-sharing features (powered by Project Astra) to all iOS and Android users this week, enabling near-real-time conversations and content streaming. Future integrations will include Google Maps directions, Google Calendar events, and Google Tasks to-do lists. Deep Research, Gemini’s AI agent, will soon support private PDF and image uploads.

Benchmarks shared by Google Imagen 4

Imagen 4, the latest AI image generator, boasts faster performance than Imagen 3, with a forthcoming variant promising up to 10x speed increase. It excels at rendering fine details (e.g., fabrics, water droplets) in both photorealistic and abstract styles up to 2K resolution. Both Veo 3 and Imagen 4 will power Flow, aiding filmmaking.

Stich

Stitch, an AI tool for designing web and mobile app front-ends, generates UI elements and code from text or image prompts, offering customization options despite initial limitations. Access to Jules, an AI agent for debugging code, has been expanded.

Project Mariner

Project Mariner, Google’s experimental AI agent, has been significantly updated to handle nearly a dozen concurrent tasks and is now available to users. It can perform actions like purchasing tickets or groceries without requiring users to visit third-party websites.

Video: Google

Project Astra

Project Astra, a low-latency multimodal AI experience, will enhance Search, Gemini, and third-party products. Collaborations with Samsung and Warby Parker are underway for Project Astra glasses, though no launch date is set.

Google Search

AI Mode, an experimental Google Search feature for complex queries, rolls out to U.S. users this week, supporting sports, finance, and “try it on” apparel features. Search Live, coming this summer, will enable real-time camera-based queries. Gmail is the first app to integrate personalized context.

Beam

Beam (formerly Starline), a 3D teleconferencing system, offers “near-perfect” millimeter-level head tracking and 60fps video streaming, along with AI-powered real-time speech translation in Google Meet.

Gemma 3n

Additional updates include Gemini’s integration into Chrome, offering an AI browsing assistant; Gemma 3n, a model for seamless performance across devices (available in preview); and numerous AI Workspace features for Gmail (personalized smart replies, inbox cleaning), Google Docs, and Google Vids (content creation/editing tools).

Wear OS 6

Wear OS 6 introduces a unified font for tiles, dynamic theming for Pixel Watches, and a design reference platform with guidelines and Figma files for developers.

Google Play

Google Play enhancements for Android developers include tools for managing subscriptions (with multi-product checkout and add-ons), topic pages for specific interests, audio samples, and streamlined checkout processes. Developers also gain dedicated testing/release pages and the ability to halt problematic live app releases.

Android Studio incorporates AI features like “Journeys” (agentic AI with Gemini 2.5 Pro) and “Agent Mode” for intricate development tasks. An enhanced “crash insights” feature analyzes source code to identify and suggest fixes for app crashes.

Featured image credit