Gemini Spark is now Task: release date, plans and price

In brief

Gemini Spark is Google's cloud AI agent, announced on 19 May 2026 and now called Task. As of November 2026 it needs Google AI Pro ($19.99 a month in the US) or Ultra, runs up to 15 tasks at once, and asks before it buys, sends or submits anything.

What is Gemini Spark, when was it released, and why is it called Task now?

Gemini Spark is Google's personal AI agent: you hand Gemini a job and it works through it in the cloud while your laptop is closed. Google announced it on 19 May 2026 with the line "rolling out today to trusted testers and in Beta to Google AI Ultra subscribers aged 18 and over in the United States in the coming weeks," per Gemini's own release notes. It ran on Gemini 3.5 and Google's Antigravity harness.

A rainbow glass four-point spark on a charcoal plinth, a terracotta toggle switch with its lever thrown on the middle plinth, and a chrome alarm clock on the right plinth, under a warm rim light.

Google changed the name, not the job. Its help page for tasks answers "Where did Spark go?" by saying the team "renamed it Task to make it easier to understand," and that existing tasks and schedules carry over. 9to5Google reported on 12 October 2026 that Chat and Task buttons now sit above the prompt box, the Spark entry left the side menu, and "Gemini Spark Settings" became "Task settings." Google's help page gives no rename date and we found no separate announcement, so 12 October rests on that one press report.

Spark became a Task button within five months of launch
GoogleSpark announced at I/OSpark on Mac9to5GoogleTask buttons seen

Google's launch posts and release notes still say Spark. If you open the app and find no Spark tab, you haven't lost the feature. Look for the Task button.

Is Gemini Spark free? Which plan you need and what it costs

You need Google AI Pro or Google AI Ultra, and that is the only plan rule Google's help page states. The Free and Google AI Plus plans aren't named. One catch: Google's US plans page names "Tasks in Gemini" only on the two Ultra cards, as first access in select countries, so confirm in your own account before paying for Pro. The table is compiled from that plans page and the Tasks help page, both read in November 2026.

PlanUS price a monthTasks in GeminiUsage on the plan card
Free$0NoBaseline usage
Google AI Plus$4.99Not named on either page2x Free's usage
Google AI Pro$19.99Yes, per the help page4x Free's usage
Google AI Ultra 5x$99.99Yes5x Pro's usage limits
Google AI Ultra 20x$199.99Yes20x Pro's usage limits

The help page adds four conditions. You must be 18 or over, sign in with a personal Google account (work and school accounts aren't supported yet), keep Keep Activity on, and live outside the European Economic Area, Nigeria, Switzerland and the United Kingdom. Tasks work in the Gemini mobile app, the Gemini app for Mac and at gemini.google.com, and nowhere else.

Usage is compute-based. Google says higher subscriptions raise the limits and that Gemini warns you when you're close. There is no per-month task count. If you downgrade or cancel, in-progress tasks finish and schedules pause, they aren't deleted. Google AI Pro limits breaks down what the plan's allowance covers.

How do you use Gemini Spark, step by step?

You click Task below the text box, describe the job, and submit it. These steps are from Google's help page for tasks as of November 2026, and the menu names are Google's.

  1. Go to gemini.google.com, or open the Gemini app on your phone or Mac.
  2. Click Task below the text box and follow the on-screen instructions the first time.
  3. Describe the job and, if it should repeat, when it should run.
  4. Optionally type @ to pick a skill, or click Upload & tools to add files, a Drive file or a notebook.
  5. Click Submit.
  6. Open the progress chip at the top of the thread to watch the steps. The Files section of the work panel lists what Gemini read or changed.
  7. Type into the box to change direction mid-task. Google says progress is kept.

A prompt that fits the shape: "Every Monday at 8 a.m., read my unread newsletters in Gmail, archive the ones I never open, and send me a one-page summary of the rest in a Google Doc." It names the app, the schedule and the finished file. Gemini can produce PDF, DOCX, XLSX and PPTX files, and Markdown or plain text files open in Canvas for editing.

To pause a schedule, use the scheduled task chip or the Scheduled page on the web. To switch the feature off, open Settings, then Task settings, and turn off "Use task mode to hand off work." Doing that deletes the remote browser and code-execution data, but not your tasks, threads or files.

What can Gemini Spark do, and when does it ask first?

It can run multi-step jobs across your Google apps, the web and files, and it stops for your approval before anything hard to undo. Google's launch post gives three jobs: scanning monthly credit card statements for new or hidden subscription fees, watching a child's school emails for deadlines and sending a daily digest, and turning meeting notes from emails and chats into a Google Doc plus a follow-up draft.

The help page adds decluttering email, tracking flight and hotel prices against a budget, and cited research delivered as a PDF, spreadsheet or presentation. The connected apps are Calendar, Docs, Drive, Gmail, Keep, Sheets, Slides and Tasks, plus Contacts, Photos, YouTube, Maps and others. Google's 30 June 2026 post lists Canva, Dropbox, Instacart, OpenTable and Zillow Rentals, plus custom MCP connections. What MCP is explains that last one.

The confirmation rule is the part worth memorising. Google says Gemini always asks before it sends communications, changes or deletes data, makes purchases, completes financial transactions, submits web forms or uses Sign in with Google, even if you set "Always allow." Bulk actions on your private Google Tasks list are the one exception.

On a Mac it goes further. Gemini Spark "only has access to the files you give it permission to use," Google wrote on 30 June 2026, and the example was sorting a Downloads folder of PDFs into folders. That Mac beta was for US Ultra subscribers at launch.

What can't Gemini Spark do?

It can't be trusted to run unwatched. Google's help page opens its limits with "Gemini's ability to handle tasks is experimental" and says "Gemini is still learning and can make mistakes." Safeguards, it adds, don't guarantee protection, so Google recommends supervising.

Four limits you can plan around. First, you can't put passwords or card numbers in the thread: the task pauses at sign-in and you take over the browser. Second, the browser comes in two kinds. Chrome auto browse is desktop-only and needs your device and Chrome awake, while the remote browser that keeps working after you close your device is used only on mobile. Third, only 15 tasks run at once. Fourth, Google doesn't publish a file size or file count limit for uploads on that page, so don't assume one.

Sensitive information such as health, financial or government ID details needs your permission before Gemini shares it, again even with "Always allow" set. Regions matter too: nothing runs in the EEA, Nigeria, Switzerland or the UK as of November 2026.

How does Gemini Spark compare with ChatGPT Work and Claude Cowork?

All three run a browser for you, but only Gemini and ChatGPT keep working after you close your device, and Claude's built-in browser needs the desktop app open. The table uses each vendor's own help page, read in November 2026.

Gemini TaskChatGPT Work cloud browserClaude Cowork built-in browser
PlansGoogle AI Pro, UltraPaid plans, "excluding Free and Go"Pro, Max, Ultra; Team and Enterprise unless an owner turns it off
Runs with device offRemote browser on mobile onlyYes, tasks keep running after you close your computerNo, "Claude Desktop needs to be open and online"
Asks beforePurchases, sends, form submits, sign-inPayments and bookings, in chat; a new website by defaultNot covered on the page
Sign-inPauses, you take overPauses, secure formYou sign in as you go or import cookies
ConcurrencyUp to 15 tasksNot published on the pageNot published on the page

ChatGPT's help page says the credentials you type go straight to the remote browser, where the model can't see them. Google says nothing equivalent in the page we read, only that you enter sign-in details on the webpage yourself. How to use agent mode in ChatGPT covers the ChatGPT side in depth.

Pick Gemini Task if your mail, calendar and documents already live in Google and you pay for AI Pro. Pick ChatGPT Work if you want the browser to carry on whatever your phone does. Pick Claude Cowork if you work at a desktop and want to watch its browser in a side panel next to the task.

Whizi can't run any of the three. It doesn't run Gemini Task, ChatGPT Work or Claude Cowork, because those are each vendor's own products. What it does is put Claude, GPT and Gemini chat in one app, with web search on by default, so the reading and comparing half of the job lives in one place. Pro's first week is $1.99, and the free plan gives 10 messages a day on Whizi AI, Auto and Claude Haiku 5.5, no card. Whizi has no clicking agent, so it won't book or buy for you.

Workflow checklist
  • Look for the Task button below the text box, not a Spark tab: Google renamed Spark in October 2026
  • Check you have Google AI Pro or Ultra, a personal account, age 18 or over and Keep Activity on
  • Name the apps, the schedule and the finished file in the first message
  • Type passwords and card numbers only on the web page after the task pauses, never in the thread
  • Keep schedules under the 15-task limit, because a schedule is skipped when 15 tasks are running
  • Read each confirmation before you approve a purchase, send or form submit
Common questions

Frequently asked questions

Is Gemini Spark free?

No. Google's help page says you need a Google AI Pro or Google AI Ultra subscription to use tasks. As of November 2026 AI Pro costs $19.99 a month in the US and Ultra costs $99.99 or $199.99. The Free plan has no Tasks access, and the help page doesn't name Google AI Plus.

Is Gemini Spark available in the UK or Europe?

No. Google's help page says tasks aren't available in the European Economic Area, Nigeria, Switzerland or the United Kingdom as of November 2026. Google says access is expanding over time, so check the help page for your country before paying.

Where did Gemini Spark go in the Gemini app?

It became the Task button. Google's help page says Spark was renamed Task, and 9to5Google reported on 12 October 2026 that Chat and Task buttons now sit above the prompt box. Your existing tasks and schedules stay in the sidebar. The rollout reached Android first, so web and iOS may lag.

How is Gemini Spark different from Claude Cowork?

Gemini runs jobs in Google's cloud and keeps a remote browser going after you close your phone, while Claude Cowork's built-in browser needs Claude Desktop open and online. Gemini needs Google AI Pro or Ultra. Claude's built-in browser is on Pro, Max and Ultra, and on Team and Enterprise unless an owner turns it off.

When was Gemini Spark released?

Google announced Gemini Spark on 19 May 2026 and began rolling it out that day to trusted testers and, in Beta, to Google AI Ultra subscribers aged 18 and over in the United States. Google's help page now lists Google AI Pro or Ultra. It doesn't say when Pro was added.

Can you download Gemini Spark?

Google's help page lists three places tasks work as of November 2026: the Gemini mobile app, gemini.google.com and the Gemini app for Mac. Google's 30 June 2026 post put the Mac beta at gemini.google/mac for Ultra subscribers.

What are skills in Gemini Spark?

Skills are reusable instructions with additional context. Google's help page says you type @ in a task and pick a skill, and that a task can use several skills along with schedules, which tell Gemini when to run.