Google I/O 2026 explained: Gemini 3.5 Flash, Spark and Android XR glasses

Google’s annual developer conference wrapped on Tuesday in Mountain View, California, after two days that delivered the company’s most ambitious AI roadmap yet. Chief executive Sundar Pichai opened the keynote by declaring the start of “the agentic Gemini era,” setting the tone for a showcase of a faster and cheaper flagship model, always-on personal AI agents, and the first consumer hardware running Android XR — Google’s operating system for eyewear that layers digital information onto the wearer’s view.

“You no longer have to trade quality for latency.”
— Sundar Pichai, Google CEO

The scale Pichai shared upfront tells the story of how quickly Gemini has caught on. The Gemini mobile app now counts more than 900 million monthly active users, more than doubling from 400 million at I/O 2025. Across its surfaces Google is now processing over 3.2 quadrillion tokens per month — a sevenfold jump from last year — and AI Overviews in Search reach 2.5 billion monthly users. Capital expenditure is tracking toward $US180 billion to $US190 billion ($273–288 billion) this year, up from $US31 billion in 2022.

Gemini 3.5 Flash: speed and cost are the story

Gemini 3.5 Flash was the headline model. Google says the frontier-class system outputs tokens four times faster than rival models at less than half the cost. Onstage demos showed it coding, reasoning through multi-step problems, and handling agentic workflows — tasks where the AI plans and executes across apps instead of just answering questions.

For Australian enterprises and startups, the cost numbers matter most. VentureBeat reported that Google estimates the model could cut enterprise AI spending by more than $US1 billion a year if workloads shift to Flash. The model is available now through Google AI Studio and the Gemini API — same access for Australian developers as their US counterparts. A larger Gemini 3.5 Pro model got a preview but no ship date. Pichai said it would arrive next month, which prompted speculation from Business Insider that safety testing held it back.

Gemini Spark: Google’s answer to the always-on agent race

Gemini Spark is a 24/7 personal AI agent that runs continuously on Google Cloud and can act on a user’s behalf — drafting emails, building study guides, watching credit card statements, and summarising the day’s notifications. It goes head-to-head with OpenClaw, the always-on agent that has been one of the most talked-about consumer AI products of 2026.

The announcement also raised the privacy questions that come with any AI product that needs persistent access to personal data. As The Verge noted, Google’s AI future “demands trust — and consumers’ personal data.” Google said Spark works within the user’s existing Google Account permissions and includes controls over which data streams it can touch, but the company stopped short of offering a detailed privacy model onstage. Fast Company argued that Google’s AI strategy is “finally coming into focus” — the search giant no longer just organises the world’s information. It wants AI to increasingly operate on the user’s behalf.

Gemini Omni and the new AI Ultra pricing

Google also showed Gemini Omni, describing it as a “world model” that can generate video from any input — text, image, audio or video. The demo drew applause but Google did not give a ship date. The company framed it as a research-direction signal, not an imminent product. What was concrete: Google cut the price of its top-tier AI Ultra subscription to $US100 ($152) per month, down from $US250. The drop makes Google’s premium AI tier look more competitive against OpenAI’s pricing and shows the subscription price war is picking up speed.

Android XR glasses are real and launching this year

Android XR smart glasses generated the most press. Built with Samsung and designed by Warby Parker and Gentle Monster, the glasses come in two styles — a sportier “disruptive” frame and a classic “timeless” design. Google also showed a reference design from Xreal that it is positioning as a developer kit. All three are scheduled to ship in the US in the northern autumn of 2026.

The glasses run Android XR with Gemini integrated throughout, enabling real-time translation, navigation prompts overlaid on the wearer’s view, and contextual information pulled up by voice. Shahram Izadi, Google’s VP and GM of XR, wrote in a blog post that the aim is “intelligent eyewear” — not a headset that isolates the wearer. Hands-on reports from WIRED and The Verge were cautiously positive. The frames feel closer to normal glasses than earlier efforts from Meta and Ray-Ban, though battery life, Australian pricing, and local availability are still unknown.

For Australian consumers, the biggest open question is timing. Google has not confirmed whether the US autumn launch includes Australia, though the Android XR platform supports global developers from day one. ZDNet Australia’s hands-on described the reference design as “limit-testing” enough to “scare Meta and Apple.”

What this means for Australian developers

The most practical announcements for Australian developers landed on the tools side. AI Studio now has a native Android app that supports “vibe-coding” — describing an app idea in natural language and having Gemini write the code. Antigravity 2.0, Google’s CLI for AI-assisted development, adds WebMCP standards support. That makes it easier to connect AI models to external tools and data sources without writing custom integrations.

Together these changes lower the barrier for Australian developers building for the Gemini ecosystem, even without deep Android platform experience. Gemini 3.5 Flash at half the cost of comparable models also shifts the economics for local startups working on AI-native products.

What to watch next

Google I/O 2026 showed a company moving past bolting AI features onto existing products. The direction is an AI-first platform: models, operating system, and hardware developed as a single stack. Gemini 3.5 Pro is expected in June. The Android XR glasses launch follows in the Australian spring. Apple’s WWDC in June and Meta’s next hardware move will determine how this race plays out. For now, Google has placed the most integrated AI-hardware bet of any major platform company.