ChatGPT Work vs Claude Cowork vs Gemini Spark, which agent actually gets work done?

Three of the biggest names in AI shipped a "hand off your work to an agent" product within months of each other. That's not a coincidence — it's a race. OpenAI, Anthropic, and Google have all decided that the next battleground isn't chat quality, it's whether the AI can actually finish the task while you go make coffee.

If you've tried to figure out which one deserves your subscription money, you've probably noticed the marketing pages all sound identical: "hand off a task," "runs in the background," "review the finished work." So this guide breaks down what ChatGPT Work, Claude Cowork, and Gemini Spark actually do differently, what they cost, and which one fits your workflow.

What exactly are these three agents anyway?

All three products share the same basic pitch — give the AI a goal instead of a prompt, and it plans, executes, and delivers a finished output instead of a chat reply.

ChatGPT Work is OpenAI's answer. ChatGPT Work is a new agentic mode inside ChatGPT that sits alongside the familiar Chat mode and the developer-focused Codex mode, connecting to tools like Slack, Gmail, Google Drive, and Salesforce through a Plugins directory, and staying with a multi-step project for hours to produce a finished spreadsheet, slide deck, report, or small web app. It's designed to help create documents, spreadsheets, presentations and web applications, and it's powered by GPT-5.6, OpenAI's latest model.

Claude Cowork got there first. Launched in January, it automates multistep workflows for personal organization and scheduling by pulling from users' files, folders and applications without the need for constant prompting or direction. You can find the official Claude Cowork product page and the Claude Help Center guide if you want to see it straight from Anthropic.

Gemini Spark is the newest entrant. Google announced a new agentic personal assistant called Gemini Spark at Google I/O, built from Gemini base models and an agentic harness from Google Antigravity, with Alphabet CEO Sundar Pichai describing Spark as the next evolution of smart digital assistants, using agentic AI to take on long-horizon tasks with minimal oversight. Google's own announcement post has the full rundown.

How is each agent actually built differently?

This is where the three start to diverge in interesting ways.

Spark's whole architecture is designed around never needing your device on. It runs on dedicated virtual machines on Google Cloud seamlessly, so you don't need to keep your laptop open to make sure it's running. That's a direct shot at how agentic tools used to work.

Claude Cowork started as the opposite — a desktop-tethered tool — before catching up. Claude Cowork launched as a desktop app in January, but it's now available on web and mobile for Max subscribers, meaning users can start a task from their desk, get status updates on their phone, and pick up the finished output later, even if their laptop is closed. Anthropic also folded a persistent chat-based version into Slack. Anthropic recently launched Claude Tag, an always-on Claude that lives in Slack and acts as an AI teammate.

ChatGPT Work leans hardest into the "finished deliverable" framing. It takes an outcome, gathers information across your apps and workflows, and stays with complex projects for hours by breaking them into smaller steps and completing them independently — and the output isn't chat. It also comes bundled with a rebuilt browsing capability. Its built-in browser gathers information from websites to build those deliverables.

Which agent wins on meeting transcripts and action plans?

Run the same test across all three — feed in a meeting transcript and ask for a summary, decisions, owners, deadlines, and next steps — and the differences come down to where each tool's strengths naturally sit.

Gemini Spark has an edge if your meeting notes live inside Google Workspace already. You can teach it to perform tasks such as creating a list of critical deadlines in your Gmail and sending it to you, or writing up a summary of ongoing updates in lengthy email threads, and you can ask Spark to look at meeting notes in your chats and emails and create polished reports in Google Docs, as well as draft an email that you can send along with that report.

Claude Cowork tends to produce the most usable "first draft" style output because of how it treats files. In Cowork, Claude has permission to read, edit, and create files in folders you specify — so it can actually complete tasks rather than just describe how to do them. Anthropic's own example is basically the meeting-prep use case in miniature: "Set Monday's client prep for 6 am: Claude works through the email threads, transcripts, and recent news, builds the briefing doc, and leaves the follow-up email drafted but unsent."

ChatGPT Work is the one most likely to hand you something polished enough to send without edits, since it's explicitly built around finished artifacts rather than research notes. It's an autonomous agent that does not hand the work back to you — it accepts a goal, connects to your apps and files, breaks the job into steps, runs independently for hours, and returns a finished artifact: a completed spreadsheet, a slide deck, a document, or an interactive web app.

What can each agent actually do beyond a single task?

The real differentiator between these tools isn't the one-off demo, it's what they do when you're not watching.

Spark is built for recurring, hands-off jobs. It's possible to set it to do recurring tasks, like making it spot hidden fees in credit card bills every month, or automatically parse monthly credit card statements to flag new or hidden subscription fees. It's also expanding past Google's own apps. In addition to Workspace apps, it will be able to link to Canva, OpenTable and Instacart, with more partner apps coming in the following weeks.

Claude Cowork leans into everyday admin cleanup. Some of the use cases Anthropic showcased include reorganizing downloads, turning receipt screenshots into expense spreadsheets, and producing first drafts from notes across a user's desktop. And Anthropic's own usage data backs up that this is where people actually use it. Based on a sample of 1.2 million anonymized Cowork sessions, 33.4% of customers used Cowork for business process and operations, including pulling scattered updates into a single report, building onboarding checklists, and reconciling spreadsheets.

ChatGPT Work is aimed squarely at knowledge-work output, not personal admin. It's designed to help create documents, spreadsheets, presentations and web applications.

How much does each one actually cost?

Pricing is where the three products look most different, and it matters more than the demo videos suggest.

ChatGPT Work is the cheapest way in, because there's no separate fee. It is not a new subscription tier — it is bundled into Plus, Pro, Business and Enterprise plans at no extra list price. Access is rolling out in waves though. It's rolling out to paid plans except Free and Go, in all supported regions, with Pro, Pro Lite, Enterprise and Edu getting access first, and Plus and Business following over the coming days.

Claude Cowork sits behind Anthropic's most expensive tier. The tool is currently available as a "research preview" only to Max subscribers on $100 or $200 per month plans.

Gemini Spark is the priciest entry point of the three. Gemini Spark is currently only available for Google AI Ultra subscribers, meaning you'll need to spend at least $100 per month, which comes with up to 20TB of cloud storage — a 30TB plan with higher usage limits is $199.99 per month.

Which agent should you actually pick?

There's no universal winner here — the right pick depends entirely on where your work already lives.

If your team runs on Gmail, Docs, and Sheets, Gemini Spark's zero-setup integration is hard to beat, since users can email Spark directly through a dedicated Gmail address and the agent can pull context from Gmail, Google Docs, and other Workspace applications without requiring manual setup, giving Google a structural advantage over rivals whose agents must rely on third-party integrations.

If you want the most polished, ready-to-send deliverables with the least budget friction, ChatGPT Work makes sense simply because it's already included in plans many people already pay for.

If your work is messier — scattered files, receipts, half-finished drafts across a desktop — Claude Cowork's file-first approach and cross-device handoff tend to fit best, especially now that it can work in the background, tag along across devices, and request human input when a decision pops up only the user can make.

How do you actually get started?

Each tool has a slightly different on-ramp, so here's the fastest path into each one.

For ChatGPT Work: Open chat.openai.com, check OpenAI's workspace agents announcement for plan eligibility, and switch your mode from Chat to Work using the mode selector in the desktop or web app once it appears on your plan.

For Claude Cowork: Head to claude.ai or the Claude Cowork product page, and if you're on a Max plan, look for Cowork next to Chat. Switch to Claude Cowork when you want to hand off a task — find it next to Chat. Anthropic's get-started guide walks through your first session.

For Gemini Spark: You'll need a Google AI Ultra subscription. Once you sign up, you can try Spark via gemini.google or in the Gemini app.

What's the real takeaway here?

None of these three agents is objectively "better" — they're optimized for different lives. Spark bets on you already living inside Google's ecosystem. Cowork bets on messy, file-heavy, cross-device work. ChatGPT Work bets on delivering something polished enough to send without a second look.

The smarter move isn't picking a winner — it's noticing which one of these already matches how you work today, and testing that one first before you pay for the other two just to compare demos.