I/O 2025
43 announcements tracked. Reality Score: 75% shipped substantially as promised.
Shipped 28
Agent2Agent (A2A) interoperability protocol
Open protocol for AI agents with 50+ launch partners (Atlassian, Box, Salesforce, SAP, etc.)
Shipped as open spec. Donated to the Linux Foundation under the Agentic AI Foundation; reached v1.2 with signed agent cards. ~150 organizations reportedly using it in production by Cloud Next 2026.
Initial announcement was actually April 2025 at Cloud Next; reinforced at I/O.
Agentic checkout in Google Shopping/AI Mode
Price tracking with budget-based agentic purchasing via Google Pay
Rolled out on Google Search and AI Mode in the US in November 2025, with merchants including Wayfair, Chewy, Quince and select Shopify stores. Canada/Australia/UK expansion via Universal Commerce Protocol underway in 2026.
AI Mode in Google Search (US rollout)
Rolling out to all US users starting at I/O 2025; powered by a custom Gemini 2.5
Rolled out broadly in the US within weeks of I/O, expanded globally (English) by August 2025, and by I/O 2026 Google said AI Mode had crossed 1 billion monthly users. AI Overviews reached 48% of queries by March 2026.
Arguably the biggest ship of I/O 2025; Search box itself was rebuilt around AI Mode at I/O 2026.
Android XR platform / Samsung 'Project Moohan' headset
Samsung's Project Moohan first Android XR headset coming 'later this year'
Samsung Galaxy XR (the productized Moohan) launched at a 'Worlds Wide Open' event October 21, 2025; on sale in Korea Oct 13 and globally after. Android XR SDK Developer Preview 3 followed in December 2025.
Deep Research with PDF/image upload and Drive/Gmail sources
PDF/image upload now; custom Drive/Gmail source selection 'coming soon'
Upload shipped at I/O. Drive and Gmail source selection rolled out into Deep Research over summer 2025. Deep Research expanded into AI Mode in Labs as 'Deep Search'.
Gemini 2.5 Pro and 2.5 Flash GA
2.5 Flash out of preview; 2.5 Pro generally available in AI Studio/Vertex AI in early June
Both 2.5 Pro and 2.5 Flash hit GA on June 17, 2025, on schedule. Google also surprise-launched 2.5 Flash-Lite that day. Family later extended with Deep Think and Gemini 3 by late 2025.
Gemini Code Assist for individuals and GitHub GA
Generally available, powered by Gemini 2.5, with a free tier
Free tier shipped on schedule with very generous limits (180k completions/mo, 240 chats/day). Agent Mode added in June, GA in VS Code/IntelliJ Aug 21, 2025.
Gemini in Chrome (desktop)
Gemini integrated into Chrome on Windows/macOS for Pro and Ultra subscribers in the US
Rolled out to Pro/Ultra subscribers around launch. Expanded to all US Chrome users (free tier) starting September 18, 2025. Expanded to India, Canada, New Zealand in March 2026. Upgraded to Gemini 3 with full agentic mode at I/O 2026.
Gemini Live with Maps, Calendar, Tasks, Keep
Integration with Google Maps, Calendar, Tasks, Keep for mid-conversation actions; 'coming weeks'
Limited release began late June 2025; wide rollout on Android and iOS in August 2025, both as promised. Camera/screen sharing on iOS shipped on day one as announced.
Gemma 3n on-device multimodal model
Fast multimodal Gemma model for phones, laptops, tablets, supporting audio/text/image/video
Preview at I/O, full release June 26, 2025 as E2B and E4B sizes via Hugging Face, Kaggle, Vertex AI, AI Studio. Compatible with Transformers, llama.cpp, MLX, etc.
Gmail Personalized Smart Replies
Smart Replies personalized using your past email tone and Drive files; rolling out 'later this year' on paid Workspace/AI Premium
Labs alpha in July 2025, with broader Workspace/AI Pro rollout through Q3-Q4 2025. Shipped roughly on its 'later this year' commitment, paid-only as stated.
Google Flow AI filmmaking tool
AI filmmaking tool built on Veo, Imagen and Gemini; available at launch for Pro/Ultra subscribers in the US
Launched on schedule May 20, 2025 to AI Pro/Ultra. Got Veo 3.1 upgrade in late 2025. Rolled out to Workspace tiers from January 2026, with dedicated mobile apps unveiled at I/O 2026.
Google Meet real-time speech translation
Near real-time speech translation in Google Meet, beta from launch day, voice-preserving
Beta opened May 20, 2025 for AI Pro/Ultra users, English-Spanish only. Italian/German/Portuguese added later in 2025. Reached GA for Workspace businesses in February 2026.
Slow language ramp-up, but technically shipped on day one as promised.
Google Vids with AI Avatars
Google Vids available to AI Pro and Ultra users; AI Avatars from script uploads
AI Avatars hit GA in Vids on August 27, 2025; upgraded to Veo 3.1-powered avatars in December 2025 with better lip sync and 12 cartoon styles. Consumer version of Vids also launched.
Imagen 4 (with Ultra and Fast variants)
Available today in Gemini app; Imagen 4 Fast 'up to 10x faster' coming soon; 2K resolution
Imagen 4 launched in the Gemini app at I/O. Imagen 4 family (Standard, Ultra, Fast) reached GA in the Gemini API and AI Studio on August 15, 2025, with Fast priced at $0.02/image.
Ironwood (TPU v7)
Reinforced at I/O after Cloud Next April 2025 reveal; 7th-gen TPU with 10x peak perf vs TPU v5p
Ironwood (TPU7x) reached GA on Google Cloud later in 2025 in 256-chip and 9,216-chip pods. Anthropic publicly committed to up to 1 million TPUs.
Big competitive shot at Nvidia Blackwell.
Jules autonomous coding agent
Coding agent in public beta worldwide, no waitlist; reads code, writes tests, fixes bugs, bumps deps
Public beta from I/O day. Exited beta August 6, 2025 with a free tier and paid Pro ($124.99) / Ultra ($199.99) plans, powered by Gemini 2.5 Pro. Now has an API.
Notable price tags for Pro/Ultra — closer to Devin-tier than to Copilot.
Live API with native audio dialogue
Preview of audio-visual input and native audio dialogue in the Gemini Live API
Live API native-audio preview shipped to developers. Computer-use (Mariner-derived) capability launched in the Gemini API and Vertex AI in summer 2025 as promised.
MedGemma medical AI model
Multimodal medical model for health app development; 'available now' as part of Health AI Developer Foundations
MedGemma 4B + 27B released on May 20, 2025 via Hugging Face and Vertex AI. Expanded with MedGemma 27B Multimodal and MedSigLIP in July 2025. MedGemma 1.5 followed January 2026.
NotebookLM mobile app (iOS/Android)
Standalone mobile apps with Audio Overviews, background playback, share-to-add sources
Shipped on May 19–20, 2025 right before I/O. Subsequent additions: customizable Audio Overview length, then full Video Overviews in July 2025.
NotebookLM Video Overviews
Convert PDFs, docs and images into narrated videos; 'coming soon'
Standard narrated-slideshow Video Overviews launched July 29, 2025. Cinematic Video Overviews (animated, Veo-powered) followed in March 2026, but locked to the $249.99 Ultra plan and English-only.
Personal Health Coach in Fitbit (Gemini-powered)
Conversational AI tutor/coach concept teased at I/O (Project Astra-style), with health coach for Fitbit teed up later
Officially unveiled at Made by Google in August 2025; public preview launched October 27, 2025 for Premium Android users in the US. iOS rollout and medical-records integration followed in 2026.
Search Live (voice + camera in AI Mode)
Real-time camera and voice interaction with Search, 'coming this summer'
Voice-only Search Live launched June 2025 in Labs; camera (video) input added July 29, 2025; full launch in the Google app without Labs opt-in on September 24, 2025; global expansion to 200+ countries in March 2026.
Stitch AI UI design tool
AI-powered UI design generation for desktop and mobile from natural language
Launched as a Google Labs experiment at I/O 2025 with single-screen generation. Major Stitch 2.0 update in March 2026 added multi-screen generation, infinite canvas, voice input, and code export in 7 frameworks. Still free Labs experiment.
SynthID Detector portal
Verification portal to identify AI-generated content via SynthID watermarks
Began rolling out to early testers (journalists, researchers, media pros) at I/O via waitlist. A unified, broader detector landed alongside Gemini 3 Pro in November 2025.
Veo 3 video generation with native audio
Available now in Gemini app for Ultra subscribers and Vertex AI
Shipped at I/O for Ultra. Veo 3.1 followed in October 2025 with longer clips and better audio; Veo 3.1 Lite launched March 2026. Free Veo access landed in Gemini for many users via Google Labs.
Probably the most-cited 'AI slop' enabler of the year.
Virtual Try-On for apparel in Search
Try-on rolling out in Search Labs to US users at launch
Available in Search Labs day-one. Reached general US rollout July 2025 across Search/Shopping/Images, with billions of apparel items. Selfie-based try-on (no full body photo needed) added December 2025.
Wear OS 6 with Material 3 Expressive
Developer Preview at I/O; biggest visual refresh in years with Material 3 Expressive, 10% battery improvement
Developer Preview shipped May 2025. Stable Wear OS 6 rolled out to Pixel Watch via the October 2025 update, then to Samsung Galaxy Watch models via One UI 8 Watch.
Shipped late 1
AI Mode personal context (Gmail/Photos)
AI Mode tapping Gmail for 'tailored results' reflecting your preferences
Quietly tested for ~7 months. Branded 'Personal Intelligence' and rolled out publicly January 22, 2026 in the US (opt-in, Gmail + Photos). Expanded to ~200 countries / 98 languages at I/O 2026 without subscription.
Shipped ~8 months after announcement.
Scaled back 7
Agent Mode in Gemini app
Experimental capability where Gemini completes multi-step goals; coming soon to Google AI Ultra subscribers
Launched (as 'Gemini Agent') to US Ultra subscribers later in 2025, handling tasks like inbox triage and travel booking. Still US-only and gated to the $249.99/mo tier.
Gemini 2.5 Deep Think reasoning mode
Experimental enhanced-reasoning mode for math, code and multimodality; coming to Gemini app for Ultra subscribers
Shipped August 1, 2025 to Google AI Ultra subscribers in the Gemini app (with a daily prompt cap). Broader API access remained limited to trusted testers into 2026. Still locked to the $249.99/mo Ultra tier as of May 2026.
Keynote pitched it as Gemini's flagship reasoning capability; in practice it is paywalled behind the $249.99/mo Ultra tier with daily caps.
Google AI Ultra $249.99/mo subscription
New $249.99/mo tier with 30TB storage, YouTube Premium, early access to Veo 3, Project Mariner, Deep Think; US-only at launch, more countries 'soon'
Launched in US at I/O and expanded to 150+ countries by May 2026, but the flagship features that justified the price — Deep Think, Project Mariner, Gemini Agent, Project Genie — remain US-only or limited in most of those markets.
Subscription is sold globally, but the headline 'why pay $249.99' features are still gated to the US a year later.
Google Beam (formerly Project Starline)
AI-first 3D video calling platform with HP launching first devices 'later this year'
HP Dimension with Google Beam unveiled June 2025 at InfoComm, priced at $24,999 (license separate), shipping to select enterprise customers in US/UK/Canada/France/Germany/Japan in late 2025. Far from the consumer 'magic window' Starline pitch.
Shipped on time, but a $24,999 enterprise-only conference-room appliance is a different product category than the consumer Project Starline vision.
Lyria 2 and Lyria RealTime
Lyria 2 expanded via YouTube Shorts and Vertex AI; Lyria RealTime interactive music model via Gemini API
Lyria 2 shipped to YouTube Shorts (initially US/UK/CA/AU/NZ) and Vertex AI. Lyria RealTime available in AI Studio. Music AI Sandbox remains gated to selected artists/producers, not broadly available. Successor Lyria 3 launched later in the Gemini app, also restricted.
Models exist but the consumer-facing Music AI Sandbox is still artist-only — narrower than the broad creator pitch.
Project Astra x Aira (blind/low-vision)
Prototype with Aira to assist blind and low-vision users with daily tasks
'Aira AI Visual Interpreter powered by Project Astra' launched in the Aira Explorer app via a Trusted Tester program — free for new and existing Aira users during the pilot, still pilot-only 12 months in.
Available only inside Aira's app as a pilot, not the broad Astra-for-accessibility product the demo implied.
Project Mariner web-browsing agent
Agentic web-browsing capability for AI Mode and Gemini; available to Google AI Ultra subscribers in the US, computer-use API to Gemini/Vertex AI this summer
Available to US Ultra subscribers from I/O. Mariner-powered tasks landed in AI Mode and Agent Mode. Computer-use API shipped to Gemini/Vertex AI in summer 2025. Still US-only/Ultra-gated as of May 2026; broader international rollout pushed to late 2026.
Available, but expensive and US-only — much narrower than the universal-agent vibe of the keynote demo.
Rebranded 1
Google AI Pro (rebrand from AI Premium)
Rebrand of $19.99 AI Premium tier to 'Google AI Pro' with higher rate limits
Rebrand executed immediately at I/O. Free 12-month AI Pro for university students rolled out in US, UK, Brazil, Indonesia, Japan and ultimately 120+ countries by January 2026.
In progress 3
Project Astra / Android XR smart glasses prototype
Glasses prototype with Gemini, Samsung partnership; trusted-tester program; Gentle Monster and Warby Parker design partners; Samsung software/hardware platform for glasses, dev access later this year
Glasses did not ship in 2025 — only kept in trusted-tester hands. At I/O 2026, Google announced 'intelligent eyewear' from Gentle Monster and Warby Parker launching Fall 2026, audio-only first. Original 2025 keynote demo glasses still not on sale.
Classic Google smart-glasses demo-to-ship gap; first consumer hardware now slated for Fall 2026.
Project Astra conversational tutor
Conversational tutor with step-by-step guidance and diagrams; 'coming to Google products and Android Trusted Testers later this year'
Demoed at I/O 2025 as a research prototype. Capabilities seeded into Gemini Live, but a dedicated 'Astra Tutor' student product is only rolling out to students in 2026 — about a year after the keynote, and the original step-by-step diagram experience has not shipped as a discrete product.
Borderline vapor — only kept in 'in-progress' because Gemini Live absorbed parts of the demo.
XREAL Project Aura Android XR glasses
XREAL making 'Project Aura' Android XR glasses for developers
Project Aura formally shown at I/O 2026 with full specs (Snapdragon, 70° FOV OLED). Developer kits ship summer 2026, consumer launch by end of 2026 — about a year later than implied.
Vapor 3
Gemini Diffusion (text diffusion model)
Experimental text-diffusion research model, 5x faster than 2.5 Pro for code; trusted testers, 'broader release in coming months'
Waitlist demo opened at I/O. Still labeled experimental / trusted-testers-only as of mid-2026, with no general API availability, no pricing, and no product surface — the 'coming months' broader release never happened.
Research preview pitched as a product line; 12 months later still no public access — calling this anything other than vapor flatters Google.
SignGemma sign-language model
Open model translating American Sign Language to English text
Preview opened to LSPs, accessibility researchers, and Deaf community members via Google's AI Developer portal. Broader release was planned for Q4 2025 but never materialized — no public model card, no open weights, and no GA path 12 months on.
Announced as an 'open model' but never shipped one — still feedback-only access a year later.
Sparkify (question → animated short video)
Convert questions into animated 2-minute videos via Veo 3 and Gemini; waitlist at I/O; 'coming to Google products later this year'
12 months later still a waitlist-gated Google Labs experiment at sparkify.withgoogle.com with no public on-ramp. Capabilities were partly absorbed into NotebookLM Video Overviews and Flow, but the named Sparkify product never launched.
A year of waitlist with no public release and no roadmap update — meets the rubric's definition of vapor.