OpenAI Dots vs ChatGPT Agent and Work: What's Different? (2026)

What's the difference between OpenAI Dots and ChatGPT agent?

ChatGPT agent works when you ask. A dot works all the time. ChatGPT agent runs a web task on a virtual computer inside ChatGPT, pauses before consequential steps, and is done when the task is done. A dot, launched at OpenAI DevDay on September 29, 2026, is a persistent agent with its own cloud computer and browser that keeps working toward your goals 24/7, even after you close the chat.

If you already pay for ChatGPT, this is the confusing bit. You now have four ways to get ChatGPT to do things for you: ChatGPT agent, ChatGPT Work, scheduled tasks and dots. They overlap, they share usage limits, and they sit on different plans.

Here's the plain version. Agent is a task you hand over. Work is a bigger multi-step project across apps and files that keeps going while you're offline. Scheduled tasks are prompts that run at a set time. A dot is a named, always-on teammate that remembers how you like things done, looks for work on its own (read-only) and talks to you in ChatGPT, Slack or Microsoft Teams.

The rest of this page compares them on what matters for your wallet and your week: what each one does, which plan gets what, how usage limits work, and which one you actually need. Everything here is as of launch day, September 29, 2026.

How did OpenAI get from Operator to ChatGPT agent, Work and dots?

Each step gave the agent more time, more tools and more independence. Operator was a browser agent you watched. ChatGPT agent put that browser inside ChatGPT. Work stretched tasks across apps and hours. Dots made the agent permanent.

Operator was announced in January 2025 as a research preview for US Pro users. It was folded into ChatGPT agent and retired as a standalone product at the end of August 2025. ChatGPT Work arrived in July 2026 for multi-step workflows that return sheets, slides, docs and simple web apps, launching on Pro, Enterprise and Edu first, with Plus and Business following.

Dots are the next step. OpenAI calls them "remarkably capable, always-on agents built to handle everything", powered by a new model, GPT-6 Astra. Launched the same day: ChatGPT Space, a shared workspace where several people edit the same Page and can tag a dot in to help, plus a Pro restructure into three tiers.

The pattern matters for your decision. Nothing on this list has been switched off except Operator. Agent, Work and scheduled tasks all still exist next to dots, and a dot can hand work to them.

What does a dot add over ChatGPT agent, Work and scheduled tasks?

Persistence, memory and initiative. Agent, Work and scheduled tasks all wait for you. A dot keeps a running picture of your goals, learns your standards from feedback, and goes looking for things to do.

It's always on. Each dot gets its own cloud computer and browser and works 24/7. You can open Activity View any time to see what it's doing, with tabs for In progress, Scheduled and Completed, and you can step in, pause it or reset it. Reset deletes its conversations, memories and schedules, so it's a real clean slate.

It remembers how you work. A dot learns your preferences, standards and working habits over time, and it keeps that context whichever channel you message it from. Tell it once in Slack that invoices go out in your template, and it should remember in ChatGPT too.

It's proactive, but read-only while it looks. OpenAI calls this "proactive research": the dot scans your connected apps in the background, but during that scan it can't send messages, change content or control your computer or browser. OpenAI's own example: an early tester's dot noticed he'd forgotten to invoice a publication, pulled the details from the email thread, drafted the invoice and sent it after he approved.

It lives where teams work. You can reach a dot in the ChatGPT desktop app, web and mobile app (including voice), in Slack and in Microsoft Teams. Inside OpenAI, a bug posted in Slack sets dots investigating. Texting is a limited US-only beta for Pro. Custom Rules let you allow an action, require approval or block it entirely, an auto-review system checks consequential actions, and sensitive tasks like changing passwords always stay with you.

It connects to more. Dots reach 4,000+ apps through OpenAI's plugin ecosystem, can use your own computer with permission, and can sign into supported sites with saved passwords the model never sees. OpenAI is also clear that dots can make mistakes, which is why those rules matter.

Does a dot replace ChatGPT Work, and which plan gets what?

No, a dot sits on top of Work and Codex, and uses them. Chatting with your dot doesn't count toward your ChatGPT usage limits. But when your dot runs a task in Codex or ChatGPT Work, that task counts toward your plan allowance, just as if you'd started it yourself.

That's the detail most people miss. A busy dot can burn through your Work allowance on its own schedule. For the first month, OpenAI says dot usage doesn't count toward eligible plan allowances, which makes month one the honest test: watch how much work it generates before the meter starts.

Allowances also changed at launch. Pro $200, now called Pro More, had its Codex and Work allowance cut from 20× Plus to 10× Plus, and GPT-6 Pro messages dropped from 200 to 100 a week. Existing subscribers keep their current limits for a while, then get a one-time credit. The new Pro 500 at $500/month has the highest usage plus an Ultrafast tier of up to about 300 tokens a second in Codex and Work.

Your first dot is included at no extra charge on every Pro tier and on Business Premium. More dots, extra speed and more workload will cost extra later, and OpenAI hasn't said how much. On Pro, dots are not available in the EEA, Switzerland or the UK at launch, with no timeline; Business Premium gets them in all supported ChatGPT regions. Enterprise, Edu and Healthcare get dots as a beta that's off until an admin turns it on. See our full breakdown of what dots cost for the plan-by-plan maths.

Which one do you need: ChatGPT agent, Work, scheduled tasks or a dot?

Match the tool to the shape of your work, not to the newest launch. Most ChatGPT users don't need a dot. Some really do.

Use ChatGPT agent for one-off web jobs: compare three suppliers, fill in a form, pull prices into a table. You're there, you ask, it does it, it pauses before anything consequential.

Use ChatGPT Work for bigger projects that end in a file: a market sheet, a slide deck, a doc, a simple web app. It keeps going while you're offline, then hands you the result.

Use scheduled tasks when you want the same prompt at the same time: a Monday news summary, a daily reminder in ChatGPT. They're on paid plans from Go, so they're the cheapest way to get ChatGPT working on a timer.

Get a dot if you run a lot of knowledge work across many apps, you live in Slack or Teams, and you'd pay $100 a month or more for an assistant that notices things without being asked. Developers, operators and managers drowning in threads are the obvious fit. For a fuller verdict on strengths and gaps, read our launch-week OpenAI dots review.

Can ChatGPT agent, Work or dots work in WhatsApp?

No. None of them lives in WhatsApp at launch. ChatGPT agent, Work and scheduled tasks run inside ChatGPT. Dots add Slack, Microsoft Teams and a US-only SMS beta for Pro. WhatsApp isn't a dots channel, and OpenAI hasn't given a date for other messaging apps beyond saying dots will "soon" come to more of them and even to the phone.

ChatGPT itself had a WhatsApp number, but Meta changed its WhatsApp Business API terms to ban general-purpose AI assistants from January 15, 2026, and OpenAI shut it down. On July 13, 2026, ChatGPT returned to WhatsApp only in the EEA, after an EU antitrust order against Meta. In the US, UK, India and everywhere else, there's no ChatGPT on WhatsApp and no announced return date.

That leaves a strange map. In Europe you can chat with ChatGPT on WhatsApp but can't get a dot on Pro. In the US you can get a dot but can't use ChatGPT on WhatsApp.

If your clients, suppliers, team and family all reach you on WhatsApp, that's a real gap. Every "I'll deal with it later" in a chat is a task that never reaches Slack, Teams or ChatGPT unless you copy it over by hand.

What's the best option if your work happens in WhatsApp?

A WhatsApp-native assistant like Agent Sonic. Dots are built for people whose work lives in Slack, Teams and a stack of SaaS apps. Plenty of small business owners, freelancers and team leads run their week from WhatsApp instead, and they don't need a $100-a-month cloud computer to stop forgetting things.

Sonic lives entirely inside WhatsApp. There's no app to install, no dashboard and no desktop setup: you save it as a contact and message it like a person, by text or voice note.

It handles the small, repeating follow-through that ChatGPT's agents aren't built around: "Remind me to call the accountant every Monday at 9." "Message Omar and ask if the delivery is still on for Thursday, then tell me what he says." In the team group: "Every Friday at 3, remind everyone to submit timesheets." Forward a contract PDF and ask "Anything about cancellation fees?" Or send a voice note: "Put lunch with Priya on my calendar next Tuesday at 1."

It also keeps notes and to-dos, remembers what you tell it, searches the web live, creates and edits images, works with Google Calendar, Sheets and Docs, and writes naturally in your language, whether that's English, Hebrew or Arabic, Arabizi included.

Honest limits: Sonic doesn't make purchases, doesn't have 4,000 integrations, doesn't run a cloud computer for hours of autonomous coding or research, and isn't in Slack or Teams. If that's your work, a dot is the better tool. If your work is chats, reminders, follow-ups and documents, Sonic is the better fit. Get access at tryagentsonic.com/contact.

Frequently asked questions