Google I/O 2026 Recap: The Agentic Era Unfolds

Google I/O 2026 Recap: The Agentic Era Unfolds

This morning, Google kicked off its annual developer conference I/O 2026 with the theme "The Agentic Era", unveiling a wave of groundbreaking products and updates. Below is a structured breakdown of all key announcements.

✦ Gemini 3.5 Flash

Google's latest flagship model, Gemini 3.5 Flash, now serves as the default model across the Gemini app, Search AI Mode, and Gemini API. Gemini 3.5 Pro remains in closed beta, slated for release next month.

Gemini 3.5 Flash

Benchmark Highlights

It outperforms the previous flagship Gemini 3.1 Pro across nearly all key metrics:

  • Coding (Terminal-Bench 2.1): 76.2% (vs. 70.3% for 3.1 Pro)
  • Real-World Tasks (GDPval-AA): 1656 Elo (vs. 1314 for 3.1 Pro)
  • Tool Calling (MCP Atlas): 83.6% (vs. 78.2% for 3.1 Pro)
  • Multimodal Reasoning (CharXiv): 84.2%

Benchmark Results

Speed

  • Output speed: 289 tokens/second, 4× faster than comparable frontier models.
  • Optimized for Antigravity: 12× faster than peers, critical for real-time agent tasks (planning, execution, validation).

Speed Comparison

Pricing & Context Window

  • API Pricing: $1.50 per 1M input tokens / $9.00 per 1M output tokens; cached inputs at $0.15 per 1M tokens.
  • Context window: 1 million tokens.
  • Cost efficiency: Over 50% cheaper than rival frontier models. Enterprise clients shifting 80% of workloads to 3.5 Flash could save over $1 billion annually.

Pricing Overview

Internal Adoption

  • Google engineers' daily token consumption surged from 500 billion (March) to 3+ trillion tokens/day (now) — a 6× increase in two months.
  • 375+ external clients each process over 1 trillion tokens monthly.

✦ Gemini Omni: The World Model

Gemini Omni is Google's new world model that understands and simulates real-world physics, spatial relationships, and causal logic — essentially a next-gen video generation model. It unifies Gemini's reasoning with Veo (video), Nano Banana (image), and Genie (interactive simulation) capabilities into a single multimodal system.

Gemini Omni

Core Capabilities

  • Physics simulation: Generates claymation-style animations of protein folding, accurately modeling amino acid chain dynamics from a single prompt.
  • Conversational video editing: Edits selfie videos via natural language (style changes, element additions, perspective shifts) while preserving original motion.
  • Any input → Any output: Accepts text, images, videos, or audio (any combination) and outputs multimodal content seamlessly.

Availability

  • Gemini Omni Flash: Launched today on the Gemini app, Google Flow, and YouTube Shorts (free for Shorts users; requires AI Plus/Pro/Ultra for Gemini app/Flow).
  • Gemini Omni Pro: Coming soon.

✦ Antigravity 2.0: Agent-First Development Platform

Google's agent-centric development platform expands from last year's IDE plugin into a full product line, with four offerings: desktop app (macOS/Linux/Windows), CLI, SDK, and Managed Agents (via Gemini API).

Antigravity 2.0

Desktop App: Agent Orchestration Hub

  • Core focus: Tasks, projects, conversations, and deliverables (not just code repos).
  • Key features: Multi-agent parallel orchestration, scheduled tasks, JSON hooks, project-level permissions, and slash commands (/goal, /grill-me, /schedule, /browser).

Desktop App

Demo: Build an OS in 12 Hours

  • Built a fully functional OS from scratch in 12 hours with 93 subagents, 15,000 model requests, 26 billion tokens, and under $1,000 in API costs — ran Doom live on the custom OS.

OS Demo

CLI: Replaces Gemini CLI

  • Rewritten in Go, sharing Antigravity's agent harness.
  • Retains agent skills, hooks, subagents; extensions renamed to plugins.
  • Migration note: Gemini CLI and Code Assist IDE plugin will shut down June 18; users must migrate to Antigravity CLI.

SDK + Managed Agents

  • SDK: Build and deploy custom agents to your infrastructure.
  • Managed Agents: Launch agents via a single Gemini API call, running in isolated Linux sandboxes with persistent state. Accessible via Interactions API and AI Studio; pre-built templates available.

AI Studio Updates

  • Integrated Antigravity coding agent.
  • New mobile AI Studio app: Build native Android/Kotlin/Jetpack Compose apps with an in-browser emulator and direct Play Console publishing.
  • Export to Antigravity desktop app; first two Cloud deployments free.

✦ Gemini Spark: Your Private AI Agent

Gemini Spark is a 24/7 private AI agent running on dedicated Google Cloud VMs, powered by Gemini 3.5 Flash and Antigravity Harness. Accessible via the Gemini app, email, or chat — works even when your device is off.

Gemini Spark

Key Use Cases

  • Cross-workspace automation: Orchestrates tasks across Gmail, Drive, Sheets, Slides, and Calendar.
  • Demo: Planned a block party by auto-pulling RSVPs from Gmail, building real-time Sheets trackers, generating Slides invites, retrieving HOA rules from Drive, and drafting follow-up emails.

Gemini Spark Demo

Mac Integration

  • Select files in Finder, long-press the Function key, and dictate tasks — Gemini parses PDFs/images and generates formatted emails (auto-corrects revisions like date changes).
  • Built by a small team in 100 days with 100+ features.

Pricing & Timeline

  • Beta rollout: Trusted testers this week; US AI Ultra subscribers next week.
  • New Ultra plan: $100/month; premium tier reduced from $250/month to $200/month.
  • Future: Chrome browser agent support this summer; Android Halo dedicated agent hub later this year; MCP integration launching soon (Instacart as a launch partner).

Pricing Timeline

✦ Google Search Overhaul

Google Search gets its biggest update in 25 years, powered by Gemini 3.5 Flash.

Key Stats

  • AI Mode monthly active users: 1+ billion; queries double quarterly.
  • AI Overviews monthly active users: 2.5 billion; all-time high search volume last quarter.

Search Stats

New Search Box

  • Multimodal input: Accepts text, images, files, videos, or Chrome tabs.
  • Real-time AI suggestions: Expands queries with contextual dimensions beyond autocomplete.
  • Global launch: Available today on desktop and mobile.

New Search Box

AI Overviews + AI Mode Merge

  • Unified experience: Slide directly into AI Mode from AI Overviews for follow-up questions with persistent context; more relevant citations and sources.
  • Launch: Live today globally.

Search Agents

  • 24/7 background info agents: Monitor complex criteria (e.g., biotech stocks with P/E <15, positive cash flow, low debt) and push real-time analysis when conditions are met.
  • Use cases: Real estate hunting, sneaker drops, market tracking.
  • Collaboration: Sync with Gemini Spark.
  • Launch: This summer.

Search Agents

Generative UI

  • Built-in Antigravity coding: Generates custom interactive UIs in real time for search queries.
  • Examples: 3D black hole spacetime simulators, interactive weekend planners (calendar sync, Maps, weather, driving times).
  • Features: Shareable, reusable; free for all users this summer, custom builds for subscribers.

Generative UI

✦ E-Commerce Suite: UCP + AP2 + Universal Cart

Google launches three tools to power agent-driven shopping, solving protocol, payment, and cross-platform cart challenges.

E-Commerce Suite

UCP (Universal Commerce Protocol)

  • Open-source standard: Defines a common language for agents and e-commerce systems (product research → checkout → shipping tracking).
  • Partners: Amazon, Meta, Microsoft, Salesforce, Stripe, Shopify, Walmart, etc.
  • Expansion: Extends to hotels, local delivery, and YouTube; launching in Canada, Australia, and the UK soon.

UCP

AP2 (Agent Payments Protocol)

  • Guardrails: Set hard limits (brand, product, budget) for agent purchases.
  • Secure authorization: Tamper-proof digital mandates link users, merchants, and payment providers; encrypted privacy data.
  • Transparent returns: Unified records for both parties.
  • Launch: Powered by Gemini Spark initially.

AP2

Universal Cart

  • Cross-platform cart: Syncs across Search, Gemini, YouTube, and Gmail.
  • Smart features: Price drop tracking, inventory alerts, compatibility checks (e.g., CPU/motherboard fit), Google Wallet integration, credit card reward matching.
  • Launch: US rollout on Search/Gemini app this summer; YouTube/Gmail support later.

Universal Cart

✦ Gemini App Redesign & Creative Tools

Neural Expressive

  • Full app redesign: Fluid animations, haptic feedback, new typography, and conversational UI.
  • Rich outputs: Custom cards, interactive images, timelines, and embedded videos (no more text walls).
  • Local dialect support: Global launch today on Android, iOS, and web.

Neural Expressive

Daily Brief

  • Morning digest: Auto-summarizes inbox, calendar, and tasks with grouped topics and next-step suggestions.
  • Launch: US AI Plus/Pro/Ultra subscribers today.

Gemini Omni in App

  • Video creation/editing: Generate music videos from raw footage + references + style prompts; 360° perspective shifts while preserving motion.
  • Launch: Paid subscribers today.

Google Pics

  • Workspace AI image editor: Recognizes object relationships; one-click removal, resizing, translation, and SynthID watermarking.
  • Launch: This summer.

Stitch

  • UI design tool: 100M+ UIs created globally; new voice collaboration, code export, and one-click website publishing.
  • Launch: Global rollout today.

Google Flow

  • Omni integration: 16 unique camera-angle videos from a single image.
  • New features: Flow Agent (parallel multitasking), Flow Tools (custom creative builders), Flow Music (song demos from recordings).
  • Launch: All features live today.

Docs Live

  • Voice-to-document: Brain dump ideas; auto-pulls Drive/Gmail content to draft, format, and edit docs.
  • Launch: Pro/Ultra subscribers this summer; syncs with Gmail/Keep.

✦ Smart Glasses

Android XR-powered Intelligent Eyewear launches in two variants: audio glasses (fall 2026) and display glasses (testing phase).

Smart Glasses

Partnerships

  • Hardware: Samsung
  • Design: Gentle Monster + Warby Parker (two exclusive designs unveiled)
  • Compatibility: Android/iOS

Smart Glasses Partners

Audio Glasses Capabilities

  • Private Gemini voice chat: Bone conduction for discreet conversations.
  • Hands-free navigation: Integrates Personal Intelligence for location history.
  • App automation: Demo: Orders coffee on DoorDash via voice — Gemini opens the app, selects items, and confirms the order.

Audio Glasses

Display Glasses

  • In-lens microdisplay: Uber pickup alerts, real-time translation, customizable widgets.
  • Developer ecosystem: Early builds in progress; full details later this year.

Display Glasses

✦ DeepMind: Science & Content Safety

Gemini for Science

  • Three experimental tools: Literature Insights (paper tracking), Computational Discovery (code from research goals), Hypothesis Generation (AI-driven scientific hypotheses).
  • Companion: Co-Scientist multi-agent research partner.

Gemini for Science

Climate & Weather

  • AlphaEarth Foundations: Planetary digital twin for simulating deforestation, food security, and ecosystem dynamics.
  • WeatherNext: Accurately predicted Hurricane Melissa's rapid intensification and landfall 3 days early during the 2025 hurricane season — now a NOAA official forecasting tool.

Biotech

  • Isomorphic Labs: Accelerates drug discovery via molecular interaction modeling; multiple preclinical candidates for autoimmune diseases and cancer.

Content Safety

  • SynthID: Watermarked 100 billion images/videos + 60,000 years of audio; adopted by OpenAI, Kakao, ElevenLabs.
  • Content Credentials: Verification expanded to Search/Chrome (right-click to validate AI content).
  • New tools: AI Content Detection API (third-party access), CodeMender security agent API (beta).
Back to blog

Leave a comment