Google I/O AI Highlights

Summary

This Google I/O keynote highlighted major advancements in AI across Google's ecosystem, with a focus on the Gemini model family, new agent-driven capabilities, and expanded multimodal experiences.
Key launches include Google Beam (3D video meetings), Project Astra (universal AI assistant features), advances in Gemini 2.5 Pro and Flash, and the integration of AI tools into products like Search, Gmail, Chrome, and Android XR devices.
Significant decisions included rolling out AI Mode and new subscription plans (AI Pro, AI Ultra), expanding real-time translation and personalization, and deepening industry partnerships.
The event featured live demos, new developer tools, accessibility initiatives, and updates on AI's societal impact, with numerous new products and capabilities available immediately or rolling out in coming weeks and months.

Early June – Google Cloud Team: Release 2.5 Flash for general availability.
Summer – Google Meet Team: Roll out real-time speech translation to more languages and enterprise customers.
Summer – Search Product Team: Launch personal context and complex analysis/data visualization in AI Mode.
Coming months – Shopping/Product Teams: Release new visual shopping, agentic checkout, and try-on features.
This week – Chrome Team: Begin roll-out of Gemini in Chrome for Gemini subscribers in the US.
Product Marketing – Across Teams: Announce and promote new Google AI Pro and Ultra subscription plans globally and in the US, respectively.

Gemini 2.5 Pro now leads benchmarks across intelligence, coding, and efficiency; Deep Think mode announced for advanced reasoning, available to trusted testers prior to broad launch.
Gemini 2.5 Flash updated with improved performance, security, and cost efficiency; general availability in early June.
Introduction of "thought summaries" and "thinking budgets" to give developers better insight and control over model actions and costs.
New text diffusion model (Gemini Diffusion) significantly reduces latency for tasks like math and code, with ongoing experiments for even faster Gemini models.
Jules, an asynchronous coding agent, publicly available for automated code tasks.

Project Astra capabilities (visual, context-aware AI assistant) rolling out in the Gemini app for Android/iOS and to Chrome.
Agent mode in Gemini app will handle complex, multi-step tasks like apartment searching, booking, and scheduling, initially in experimental release.
Personal context: Users can grant Gemini access to select Google app data (Gmail, Drive, Calendar, etc.) to generate highly personalized responses; available for Gmail this summer and Search AI mode this summer.
Deep research in Gemini (Canvas) now allows file uploads and soon will integrate Google Drive/Gmail content for richer research and content creation.

AI Overviews surpass 1.5B monthly users; Google Lens usage has grown 65% YoY.
AI Mode—a new AI-powered search experience—rolling out in the US, using Gemini 2.5 for advanced reasoning, personalized results, complex data analysis, and agentic tasks.
Personalization in Search via "personal context" will allow opt-in integration with users' Gmail and app data.
Integration of Project Mariner for web/agentic operations, automating routine processes like ticket buying or reservations.
Expansion in live multimodality (Search Live) with real-time camera-based Q&A and dynamic UI adaptations.
New shopping features: visual inspiration mosaics, virtual try-on with custom-trained models, agentic checkout, and price tracking/notifications, rolling out in coming months.

Imagine 4 model introduced for high-quality, typographically accurate image generation; 10x faster than prior models.
V3, the latest video model, now generates realistic video with native audio—including dialogue and sound effects.
Music AI Sandbox and Lyria 2 for music/audio generation, available for creators and enterprises.
Expansion of SynthID watermarks and new detection capabilities for generated media (image, audio, text, video).
Flow, a new creative AI tool for filmmakers, launched with combined video, image, and audio generation capabilities; available immediately for Ultra subscribers.

Google Beam (3D video communication) introduced in partnership with HP, early customer release later this year.
Android XR platform announced for headsets and glasses; Samsung's Project Muhan headset launching this year, with Gentle Monster and Warby Parker as eyewear partners.
Live on-stage demo of Android XR glasses with Gemini for real-world context-aware tasks and real-time, on-device translation.
AI for accessibility: Project Astra prototypes supporting blind and low-vision users, ongoing partnership with IRI for supervised deployment.
FireSat constellation leveraging AI for early wildfire detection in partnership with external agencies and disaster relief efforts.

Launch of Google AI Pro and AI Ultra plans: global availability for Pro, US-first for Ultra; Ultra plan includes early access to top-tier features, expanded usage, Flow, YouTube Premium, and enhanced storage.
Many new AI capabilities (Gemini Live, Imagine 4, V3, Canvas, Gemini in Chrome, etc.) available today or this week, with more rolling out in coming weeks/months or in Labs.

AI Mode rolling out in US search — To provide an advanced, AI-powered search experience and drive engagement.
Launch of Google AI Pro and Ultra subscription plans — To monetize advanced AI capabilities and support a tiered user experience.
Rollout of personal context and agentic features — To increase value and personalization across Google’s core products, with user opt-in and privacy controls.

Exact global rollout dates for AI Ultra plan and Flow tool.
Further information on platform safety evaluations for Gemini 2.5 Pro Deep Think.
Timing for broader language support and enterprise availability of real-time translation in Google Meet.
Additional details on developer tool access and SDK compatibility timelines for new agentic/multimodal features.
Clarity on when XR glasses will be available for developers beyond trusted testers.