Google didn’t hold back at I/O 2026 on May 19-20. Over 100 announcements, a new flagship model, and an AI agent that’s supposed to work for you around the clock. Here’s what actually matters.
Gemini 3.5 Flash: Fast and Smart
The new flagship is called Gemini 3.5 Flash, and it’s available immediately — in the Gemini chat, via the API, and in Google AI Studio. Google claims it’s four times faster than comparable frontier models while matching or exceeding their quality. It scores 76.2% on Terminal-Bench 2.1, putting it in striking distance of Claude Opus 4.7 and GPT-5.5.
What’s notable: Flash was always the little sibling — fast but not particularly capable. With 3.5 Flash, that distinction disappears. Google is pushing a single model that does both.
Gemini Spark: The Agent That Never Sleeps
Perhaps the most important announcement is Gemini Spark. A personal AI agent running on Google Cloud VMs — even when your laptop is closed. Spark can autonomously navigate Google Workspace, use third-party apps, and research on the web.
This is Google’s answer to OpenAI’s Operator and Anthropic’s Managed Agents. The difference: Spark runs persistently in the cloud, not in your browser. Initially available as a beta for AI Ultra subscribers.
Antigravity 2.0: The Developer Platform
For developers, there’s Antigravity 2.0, Google’s agent-first platform. In a live demo, they built an entire operating system — and then ran Doom on it. Managed Agents in the Gemini API let you provision a Linux sandbox with a single API call where the agent can work autonomously.
Price Change: AI Ultra Starting at $100
Google is dropping the entry price for AI Ultra from $250 to $100 per month. That includes all Gemini models, Spark, 30 TB of Google One storage, and priority processing. The old $250 tier stays as a premium option at $200.
The Bigger Picture
Google is showing with I/O 2026 that Gemini is no longer just a chatbot. With Spark as an autonomous agent, Flash as an all-round model, and Antigravity as a developer platform, Google is positioning itself as a full-stack AI provider. Whether Spark delivers in the real world what the demo promised — that’s what the coming weeks will tell us.
Sources: