Stayed up for the full keynote—here are the highlights: Gemini Omni world model, 3.5 Flash, Android as an intelligent system, and smart glasses you keep on.
May 19, 2026, 10:00 a.m. PT—Mountain View, California.
While much of the world debated Valentine’s Day gifts, Google CEO Sundar Pichai opened with: “We are entering the Agentic Gemini Era.”
In plain terms: Gemini used to be a chat partner in a dialog box; now it’s meant to get work done for you. A decade of “AI First” finally shows up in everyday products.
The numbers that stole the show
- Gemini app MAU jumped from 400M last year to 900M+
- Monthly tokens processed: 480 trillion → 3.2 quadrillion (~7× in a year)
- AI Mode hit 1B MAU in its first year
- AI Overviews: 2.5B+ MAU
- Nano Banana image model: 50B+ images generated
CapEx: $31B (2022) → $180–190B projected (2026)—nearly 6×.
1. Gemini Omni: a “world model”—sketch a circle, get a black hole
DeepMind’s Demis Hassabis introduced a world model: any input, any output—text, image, audio, and video in; whatever you need out.
On stage: draw a circle on paper, say “turn this circle into a black hole,” and Omni returns a short clip with convincing gravitational lensing. Want glass buildings as soap bubbles instead? One sentence—and bubbles collide with physics.
Google’s line: “Gemini Omni is like Nano Banana for video.” It also models gravity, momentum, and mechanics—and feeds frontier robotics training.
Gemini Omni Flash is live in the Gemini app, Google Flow, and YouTube Shorts for Google AI Plus, Pro, and Ultra. Omni Pro follows.
2. Gemini 3.5 Flash: faster than ordering bubble tea
Gemini 3.5 Flash beats Gemini 3.1 Pro on benchmarks while outputting tokens at ~4× the speed of comparable frontier models. It’s now the default in the Gemini app and Search AI Mode. Internal dev workloads went from ~500B tokens/day to 3T+.
3. Gemini Spark: your 24/7 personal agent
Gemini Spark is a always-on personal agent on dedicated Google Cloud VMs—24/7, even when your phone is off: email triage, calendar, cross-app shopping, and more. Android’s new Halo notifications show agent progress. Rolling out to Google AI Ultra subscribers next week.
4. Search: biggest redesign in 25 years
Search is no longer keyword guessing—you describe the full task. AI infers intent and builds a generative UI with tailored layouts for video, images, and news. Ask how black holes form; get an embedded AI animation in results.
Search agents also ship in Chrome and Android as always-on “AI butlers” with background alerts.
5. Android: from OS to “intelligent system”
The Android Show preview framed Android as an intelligent system with Gemini Intelligence across phone, watch, car, and laptop.
- Cross-app workflows: shopping list → cart → delivery via power-button hold
- Rambler: cleans “um, uh, like” into polished text
- Create My Widget: natural-language home-screen widgets
- Android 17 Beta on all Tensor Pixels
- Android CLI 1.0 for agentic coding tools (Claude Code, Codex, etc.) against Android Studio
6. Android XR glasses
First live look at Jinju glasses (Google × Samsung): ~50g, Snapdragon AR1, 12MP camera, directional audio.
- Audio-only AI glasses: fall 2026
- Display glasses: end of 2026
Partners include Warby Parker and Gentle Monster. XREAL Project Aura—first glasses on Android XR—ships globally in 2026.
7. Everything else
- Search AI Inbox + Gmail chat search
- Ask YouTube: natural-language clip search and jump
- Docs Live: voice-driven docs
- Universal Cart: cross-store shopping agent
- SynthID watermarking
- Pics: AI design in Workspace
Pricing & access
| Tier | What you get |
|---|---|
| Google AI Ultra | Gemini Spark first (next week) |
| Plus / Pro / Ultra | Gemini Omni + 3.5 Flash |
| Free | Updated Search + Gemini basics |
Gemini 3.5 Pro arrives next month—Google claims ~half the price of rival flagship tiers.
Closing thought
Pichai: “We’ve said AI First for ten years; today it’s in products billions use daily.” Models aren’t islands anymore—agents are wired into Search, Chrome, Android, and glasses.
An Android that runs errands, a Spark that never sleeps, an Omni that “feels” gravity in video—the future landed at yesterday’s keynote.
(After staying up for the whole show until 4 a.m., my body agrees: this keynote was worth it.)