Google unveils Gemini 3.5 Flash, Spark, and Universal Cart at I/O 2026
Serge Bulaev
Google announced Gemini 3.5 Flash, Gemini Spark, and Universal Cart at its I/O 2026 event, with a focus on smarter AI helpers. Gemini 3.5 Flash may offer faster, more efficient responses in Search and the Gemini app, and appears to be cheaper than older models. Gemini Spark is described as a personal AI agent that keeps working in the background and connects to Google and outside services. Universal Cart lets people add products from different Google services and may help users find deals, but experts suggest it could also make merchants more dependent on Google. These updates suggest Google is moving toward having AI agents help with daily tasks and shopping.
Google has announced Gemini 3.5 Flash in official documentation and blogs, alongside reports of additional AI agent developments including Spark and Universal Cart features. These updates indicate a significant shift toward AI agents, introducing faster models, persistent personal assistants, and unified e-commerce capabilities. According to industry reports, these announcements underscore Google's strategy to evolve beyond a simple question-answering engine into a comprehensive ecosystem where AI agents autonomously handle user research, tasks, and shopping.
Agent models headline Google's AI emphasis: Gemini 3.5 Flash, default AI Mode, Gemini Spark, and Universal Cart
Google's recent AI developments center on a new generation of AI agents. Key announcements included Gemini 3.5 Flash for faster responses, Gemini Spark as a persistent personal assistant, and Universal Cart for integrated shopping, all designed to automate daily tasks and research for users.
The new flagship model, Gemini 3.5 Flash, is now available for free in the Gemini app and Google Search. According to reports, it offers "Opus 4.7-level" intelligence, generates tokens significantly faster than its predecessors, and operates more efficiently. The cited Google source shows Gemini 3 Flash at $0.50 per million input tokens and $3 per million output tokens, but caching discount details could not be verified from the original sources provided. Gemini 3.5 Flash will also serve as the default engine for AI Mode in Search. This is part of a broader redesign enabling the search box to handle richer prompts, marking the beginning of an "era of Search agents." These agents will monitor topics continuously, delivering updates directly to the results page.
Persistent personal help through Gemini Spark
Gemini Spark is a personal AI agent designed to run continuously on Google Cloud, functioning even when devices are off. It integrates natively with Google Workspace apps like Gmail and Docs and connects to third-party services using the Model Context Protocol. According to industry reports, Spark is entering a beta phase for U.S. Google AI Ultra subscribers, requiring human approval for sensitive actions like purchases. This persistent agent model is expected to offer greater reliability for complex, multi-step tasks, reducing development overhead. Industry experts see this as a move by Google to standardize background execution, tool integration, and action confirmations across its entire product lineup.
Universal Cart binds commerce surfaces together
With Google's Universal Cart, users can create a single shopping cart populated with items from Search, Gemini, YouTube, and Gmail. The AI-powered cart will automatically monitor for price drops, display price history, and send restock alerts. Integration with Google Wallet enables it to surface loyalty perks and payment rewards. This feature is built on the Universal Commerce Protocol, a shared framework for simplifying cross-platform checkout. The initial rollout is planned for the U.S. this summer via Search and the Gemini app, with YouTube and Gmail support to follow. For merchants, this unified system highlights the growing importance of maintaining accurate product feeds and competitive pricing, as Google's agents will automatically compare deals. However, analysts caution that this could also increase merchant reliance on Google for customer acquisition and transaction data.
Reported performance table
| Feature | Source metric or claim |
|---|---|
| Gemini 3.5 Flash speed | Significantly faster output tokens per second (Mashable) |
| Gemini 3.5 Flash cost | $0.50/M input, $3/M output tokens (Google source) |
| Spark subscription | Available for AI Ultra tier subscribers (Cheesecake Labs) |
| Universal Cart rollout | U.S. Search and Gemini app in summer, YouTube/Gmail later (Google Blog) |
Taken together, these announcements confirm Google's strategic pivot to agent-first services designed to operate across devices, continuously process information, and execute increasingly complex actions on behalf of users.
What exactly is Gemini 3.5 Flash and how does it compare to prior models?
Google positions Gemini 3.5 Flash as a mid-tier model that reaches what the company calls Opus 4.7-level intelligence while running significantly faster and costing substantially less than the previous frontier baseline.
Internally it becomes the default model for AI Mode inside Search and for the free tier of the Gemini app.
The cited Google source shows Gemini 3 Flash at $0.50 per million input tokens and $3 per million output tokens, though caching discount details could not be verified from original sources (reported via I/O transcript).
How does Universal Cart work across Google, Amazon, Meta and Microsoft?
Universal Cart is an AI shopping layer that keeps one persistent cart across Search, Gemini, YouTube and Gmail.
It automatically tracks price drops, stock alerts and loyalty perks, then lets you check out with Google Pay or hand the cart off to any merchant.
Behind the scenes Google is using the Universal Commerce Protocol to sync inventory and pricing data in real time, and the first wave of merchants already include Amazon, Meta Shops checkout partners and Microsoft's Dynamics 365 Commerce network (Google Shopping Blog).
When will Gemini Spark be available and what can it do?
Gemini Spark is a continuous personal agent that runs on Google Cloud VMs even when your laptop is closed.
It plugs into Gmail, Docs, Calendar and third-party tools (Canva, OpenTable, Instacart) through the open MCP standard.
According to industry reports, Spark will enter trusted-tester access and roll out to AI Ultra subscribers in the U.S. (Gradient Flow analysis).
Why is Google switching to an "agent-first UX" in Search?
According to industry reports, Search's classic query-result page is becoming an agent workspace.
Users can spawn information agents that stay alive, monitor pages and bring back answers or deals when conditions change.
Industry coverage describes this as a major redesign of the Search experience (Google Search Blog).
What should retailers prepare for with Universal Cart?
Retailers gain a single integration point to reach users across Search, YouTube and Gmail, but they also face greater pricing transparency because Google will surface competing deals automatically.
Feeds must include real-time inventory, loyalty rules and compatibility data to stay visible inside the agentic ranking layer.