Google Gemini 2026 marks the biggest transformation yet — at Google I/O on May 19, Google unveiled Gemini 3.5 Flash, Gemini Spark, and Gemini Omni, turning Gemini from a chatbot into a full AI operating system.
The event was not about a single model update. It was a full-scale repositioning of Gemini — from a chatbot you ask questions to, into an AI system that works on your behalf, 24 hours a day, across every Google product you already use.
Three announcements defined the keynote: Gemini 3.5 Flash, Gemini Spark, and Gemini Omni. Here is what each one does, who gets access, and why it matters.
Gemini 3.5 Flash: The New Default for Everything
Google’s first announcement was Gemini 3.5 Flash — the newest model in the Gemini family, and now the default model powering both the Gemini app and Google Search’s AI Mode globally.
Sundar Pichai described it simply: “You no longer have to trade quality for latency.” That is the core pitch for Flash — frontier-level intelligence delivered fast enough to power real-time search and agent tasks without noticeable delay.
On benchmarks, Gemini 3.5 Flash outperforms Gemini 3.1 Pro on challenging coding and agentic tests including Terminal-Bench 2.1 (76.2%) and MCP Atlas (83.6%). Google has also strengthened its safety guardrails — the model is less likely to generate harmful content and less likely to mistakenly refuse safe queries.
For Google Search specifically, the upgrade is significant. Gemini 3.5 Flash now powers AI Mode for all users globally, alongside what Google is calling the biggest redesign of the Search box in over 25 years — a dynamically expanding input field that anticipates user intent and helps formulate queries.
Gemini 3.5 Pro, the heavier-weight version, is being used internally at Google and is expected to reach wider availability next month.
Gemini Spark: Google’s Autonomous AI Agent
The announcement that drew the most attention at I/O 2026 was Gemini Spark — a fully autonomous AI agent that runs continuously in the background of your Google account.
Spark is not a chatbot. It does not wait for prompts. It monitors your Gmail, Calendar, Docs, and connected third-party apps via the Model Context Protocol (MCP), and takes action on your behalf under your direction.
What Gemini Spark Can Do
- Monitor background tasks continuously — apartment listings, product drops, flight price changes — and alert you when your parameters are met
- Proactive task execution — drafting emails, generating reports, scheduling, and integrating with external services including Canva and Instacart
- Agent Payments Protocol — Spark will eventually be able to book services like Uber and OpenTable and complete purchases, though it cannot spend money independently at launch
- Android Halo integration — later this year, Spark will connect with Android Halo to show agent progress at the top of your phone screen from any app, without interrupting what you are doing
Spark runs on dedicated virtual machines on Google Cloud and is available 24/7 — you do not need to keep your laptop or phone open for it to work. It is powered by Gemini 3.5 and the Google Antigravity platform, which enables long-horizon tasks across multiple steps and connected services.
Daily Brief
One of Spark’s most immediately practical features is Daily Brief, rolling out now for all paid U.S. subscribers (18+). Every morning, Gemini works overnight to analyze your inbox, calendar, and upcoming tasks, then presents a concise, actionable digest when you wake up — identifying urgent emails, upcoming meetings, reminders, and follow-up actions.
Access and Pricing
Spark is included in the new $100/month Ultra plan, which also includes higher usage limits, 20 TB of cloud storage, and a YouTube Premium subscription. Google also restructured its top tier — the previous $250/month plan dropped to $200/month with higher limits. Daily Brief rolls out to all paid subscribers including Plus and Pro plans.
Gemini Omni: Video Generation That Understands Physics
The third major announcement was Gemini Omni — a new multimodal model that generates video from any combination of text, image, audio, and video input.
What makes Omni different from existing text-to-video tools is its understanding of physical reality. The model is trained on all four data types simultaneously, giving it an understanding of forces like gravity, kinetic energy, and fluid dynamics. Videos it generates maintain consistent physics, consistent characters across scenes, and accurate continuity when you make step-by-step edits using natural language.
Key Capabilities
- Take an existing video and change what is happening in a scene using a text prompt
- Add new characters or objects to existing footage
- Apply cinematic effects — zooms, transitions, lighting changes — conversationally
- Generate AI avatars that look and sound like the user
- All videos include an imperceptible SynthID digital watermark for AI content verification
Where Omni Is Available
Gemini Omni Flash — the first model in the Omni family — is available now to Google AI Plus, Pro, and Ultra subscribers globally through the Gemini app and Google Flow. It is also available in YouTube Shorts Remix and the YouTube Create app for users 18+ at no cost.
Google says Omni will eventually support any output from any input — text, images, and audio generation are coming after the initial video launch.
The Redesigned Gemini App: Neural Expressive
Alongside the model and agent announcements, Google completely redesigned the Gemini app with a new visual language called Neural Expressive.
The updated interface includes fluid animations, brighter colors, new typography, haptic feedback, and a pill-shaped prompt box that expands dynamically. Gemini Live — Google’s real-time conversational mode — is now integrated directly into the main app experience rather than sitting as a separate feature.
The redesign is rolling out globally today across Android, iOS, and the web. Google also switched the app from daily usage caps to a compute-based system that refreshes every five hours, meaning fewer unexpected cutoffs mid-task.
What Google DeepMind’s CEO Said
After the keynote demos concluded, Google DeepMind CEO Demis Hassabis told reporters: “We’re at the foothills of the singularity.” The comment was not in the prepared presentation — it came in response to a direct question about where the technology is heading.
That framing — AI moving from answering questions to autonomously completing long-horizon tasks — was the underlying message across every announcement at I/O 2026.
Google I/O 2026 Gemini Announcements: Quick Summary
| Feature | What It Is | Availability |
|---|---|---|
| Gemini 3.5 Flash | New default model for Gemini app and Search AI Mode | Now, globally |
| Gemini Spark | Autonomous 24/7 AI agent across Gmail, Calendar, Docs | Ultra plan ($100/mo) |
| Daily Brief | Morning digest from inbox, calendar, and tasks | All paid US subscribers now |
| Gemini Omni Flash | Video generation from any input type | Plus/Pro/Ultra globally now |
| Neural Expressive | Redesigned Gemini app interface | Rolling out globally now |
| Android Halo | Agent progress overlay on Android | Later in 2026 |
Frequently Asked Questions
What is Gemini Spark? Gemini Spark is an autonomous AI agent inside the Gemini app that runs 24/7 on Google Cloud, monitoring your Gmail, Calendar, and connected apps to complete tasks and surface information proactively — without you needing to prompt it.
What is Gemini 3.5 Flash? Gemini 3.5 Flash is Google’s newest model, now the default for the Gemini app and Google Search AI Mode globally. It delivers frontier-level performance at lower latency and outperforms Gemini 3.1 Pro on key agentic and coding benchmarks.
What is Gemini Omni? Gemini Omni is a new multimodal model that generates video from text, image, audio, or video input. It understands physics and maintains consistent characters and continuity across edits. The first version — Omni Flash — is available now.
Is Gemini Spark free? No. Gemini Spark is included in the new $100/month Ultra plan. Daily Brief, however, is rolling out to all paid subscribers including the cheaper Plus and Pro tiers.
What is the Neural Expressive redesign? It is the new visual design language for the Gemini app, featuring fluid animations, haptic feedback, brighter colors, and a redesigned prompt interface — rolling out globally now across Android, iOS, and the web.
How is Gemini changing in Google Search? Gemini 3.5 Flash is now the default model for AI Mode in Search globally, alongside the biggest redesign of the Google Search box in over 25 years — an AI-powered input that expands dynamically and anticipates user intent.
Key Takeaways
- Google I/O 2026 marked a clear shift: Gemini is no longer a chatbot — it is an autonomous AI operating layer across Google’s entire product ecosystem.
- Gemini 3.5 Flash is now the default model for the Gemini app and Google Search AI Mode globally.
- Gemini Spark is Google’s answer to autonomous AI agents — a 24/7 background agent connected to Gmail, Calendar, Docs, and third-party apps via MCP.
- Gemini Omni introduces physics-aware video generation from any input type, available now to paid subscribers.
- Google restructured pricing: Ultra dropped to $100/month, the previous top tier dropped from $250 to $200/month.
- DeepMind CEO Demis Hassabis told reporters post-keynote: “We’re at the foothills of the singularity.”
Last updated: May 2026. Feature availability and rollout timelines are based on Google I/O 2026 announcements.
