Listen to this article:
What is Gemini Spark and why it's different from a classic chatbot
Gemini Spark isn't another chatbot where you type questions and it replies. Google describes it as a "24/7 personal AI agent" that runs in the background without needing an open app. While regular Gemini waits for your prompt, Spark actively monitors your digital environment and acts on the tasks you've assigned.The agent is powered by the new Gemini 3.5 Flash model, which Google called its "most powerful agentic and programming model" at the same conference. According to internal benchmarks, it surpasses the current Gemini 3.1 Pro in coding tests (Terminal-Bench 2.1: 76.2%) and agentic tasks (MCP Atlas: 83.6%). In output speed, Google claims it is up to four times faster than competing models of a similar category.
The key difference from existing assistants: Spark is deeply integrated with Google Workspace — it understands Gmail, Docs, Slides, and gradually more applications. And because it runs in the cloud, it doesn't need your phone or computer to be turned on. It processes tasks on Google's servers and you pick up the results when you return to your device.
What Spark can do — real-world examples
Google showed several scenarios during the keynote that give a good idea of what the agent is designed for:Finance and household management
Spark can go through credit card statements every month and look for hidden fees or unwanted subscriptions. It finds an item you forgot about and alerts you. Similarly, it can monitor email communication from your children's school and summarize important dates and changes.
Workflow automation
The stronger side is multi-step work processes. Google describes a scenario where Spark, on your command, goes through meeting notes scattered in chats and emails, creates a comprehensive report in Google Docs, and even drafts an accompanying email. All without your intervention — you just set the goal.
Reservations and purchases
Thanks to integration with partner apps (Canva, OpenTable, Instacart), Spark can search for available tables at a restaurant, compare prices, or create graphic designs. Google promises more partner apps in the coming weeks.
What's coming this summer
During summer 2026, Google plans to teach Spark to send emails and text messages and control a web browser. Spark will also arrive in the Gemini desktop app, from where it will be able to access files on your computer and automate workflows across desktop applications.
How much it costs and who gets access
Gemini Spark won't be free. First access goes to trusted testers this week. Starting next week, Spark will open to beta testers with a Google AI Ultra subscription in the US. Enterprise Google Workspace customers will get access "soon" through the Gemini app.
As for Google's pricing plans, the current structure is:
- Google AI Pro — $20/month. Basic access to AI features, without Spark.
- Google AI Ultra (lower tier) — $100/month. Five times the usage limit compared to Pro, 20TB of cloud storage, YouTube Premium. Newly announced today.
- Google AI Ultra (top tier) — $200/month (reduced from the original $250). Twenty times the limit compared to Pro, access to Project Genie and other experimental features.
For Czech users, it's not yet clear whether Spark will be available outside the US from the start. Google follows a familiar pattern with most AI innovations — first the US, then other English-speaking markets, then the rest of the world. Czech localization of Gemini already exists, so the medium-term outlook is positive, but no specific date for Europe has been given yet.
How Spark stacks up against the competition
Google isn't the first to come out with agentic AI. OpenAI launched Operator last year — an agent capable of controlling a web browser and making reservations or purchases. Anthropic offers Computer Use, a Claude model feature for computer control. Microsoft integrates agents into Copilot across Office 365.
Google's advantage: data. A billion people use Gmail, Calendar, Docs. Spark can see this data (with user consent) and leverage it in a way that competition without a similar user base can hardly replicate. The second advantage is the cloud architecture — Spark doesn't need to run on your device, so it processes tasks even when you're offline.
Disadvantage: price. At $100 to $200 per month, Spark isn't something the average user can afford. For comparison — ChatGPT Plus costs $20/month, Claude Pro also $20. Spark is for now a premium product for businesses and tech enthusiasts with deeper pockets.
Security and privacy — what Google promises
Google emphasizes several safety measures for Spark. The agent is fully opt-in — you must explicitly enable it and choose which apps it can access. Before "high-risk actions" such as spending money or sending emails, Spark will ask for your confirmation.
The Gemini 3.5 model powering Spark has, according to Google, strengthened safety mechanisms, including reasoning checks before generating a response. This is meant to reduce the risk of harmful or dangerous output. All content generated by Google's AI tools will be marked with a SynthID digital watermark.
What it means for Czech users and businesses
For a Czech user, Gemini Spark is currently more of a future prospect. However, if your company uses Google Workspace and operates in English, you could theoretically get into beta testing through a corporate account. For individuals in Czechia, the key factor will be how quickly Google releases Spark to the European market — and whether it will need to resolve compliance with the EU AI Act, which imposes stricter requirements on transparency and safety for agentic systems with access to personal data.
One thing is certain: agentic AI is moving from buzzword to product. And Google is going all in.
When will Gemini Spark be available in the Czech Republic?
Google has not yet announced an official date for the European market. Spark launches as a beta for Google AI Ultra subscribers in the US starting next week (May 2026). Given that Gemini already supports Czech, expansion to Europe can be expected within months, but we don't know the exact timeline. Companies using Google Workspace with an English environment may gain testing access sooner.
What's the difference between Gemini Spark and the regular Gemini assistant?
Regular Gemini is a conversational AI — it answers your questions, generates text, analyzes files. Spark, by contrast, is an agentic system: it runs in the background 24/7, actively monitors your apps (Gmail, Docs, Calendar), and independently completes assigned tasks without needing your interaction. While Gemini waits for a prompt, Spark works continuously and you just pick up the results.
Is Gemini Spark safe? What happens to my data?
Spark is fully opt-in — you must enable it yourself and choose which apps it can access. Before risky actions (payments, sending emails) it will ask for your confirmation. It runs on Gemini 3.5 models with strengthened safety mechanisms. Still, this is a system with deep access to personal data and we recommend carefully considering which apps you grant it access to. In the EU context, it will be subject to AI Act requirements for high-risk AI systems.
