OpenAI’s Dots get their own computer and work while you sleep, from $100 a month

OpenAI has given ChatGPT a helper that keeps working after you close the tab. Each of its new Dots is an always-on agent with its own cloud computer and web browser, wired into thousands of apps, and it chips away at the goals you hand it while you are busy, offline or asleep. Getting one is the hard part: the first Dot comes only with ChatGPT’s Pro and Business Premium plans, not the cheaper Plus tier, and individual subscribers in the European Economic Area, Switzerland and the UK are left out for now.

What separates a Dot from a chatbot is that it does not wait to be asked. In OpenAI’s favourite example, an early tester’s Dot spotted an article that had never been billed, drafted the invoice and sent it once the tester said yes. A developer can point one at the feedback arriving for an app and get back tested bug fixes, each with a video of the change. You talk to it like a colleague, by message or voice call in ChatGPT on desktop, web and mobile, or inside Slack and Microsoft Teams, and it writes back when it needs a decision.

Under the hood, every Dot runs on GPT-6 Astra, OpenAI’s most capable public model, and lives on a cloud machine you can open at any moment to watch it work, with more than 4,000 apps within reach through OpenAI’s plugins. Letting it use your own PC as well is optional and switched off by default. When you step away, the Dot carries on with what OpenAI calls proactive research, using read-only access to your connected apps. In that mode it can look, but it cannot send messages, edit content or take over a browser or computer.

You draw the lines yourself with Custom Rules, which decide what a Dot may do alone, what needs your approval and what it must never touch. Some jobs, such as changing a password, always stay with the human. Saved passwords let it sign in to supported websites without the model ever seeing them, an auto-review layer checks any action that could affect your accounts or share information, and OpenAI’s monitoring system can pause or stop a Dot that raises a safety flag.

Those fences exist for a reason. OpenAI recently paused its most capable models after an agent slipped out of a test sandbox, and it held back GPT-6.1 Astra after internal testing raised safety concerns. The company says plainly that Dots can still make mistakes and asks people to review consequential work before it goes out. Price is the other open question. Meta’s rival agent, Muse, is free up to a usage limit and costs from $20 a month after that, while a Dot requires a plan that starts at $100 a month and there is no free tier to try one.

Where you live decides the rest. The Pro rollout skips the European Economic Area, Switzerland and the UK, so most of Europe waits, while companies on the Business Premium plan get Dots in every region where ChatGPT is supported. Enterprise, education and healthcare workspaces can test a beta, but only after an administrator turns it on. No date has been announced for European Pro subscribers.

Dots also land in a crowded field. Google sells Gemini Spark as a round-the-clock personal agent, and the startup Instinct has just raised $1 billion at a $10 billion valuation. OpenAI’s pitch is the model inside: few rivals can put a frontier system behind an agent that never logs off.

OpenAI unveiled Dots on September 29 at its DevDay conference in San Francisco, and the rollout to Pro and Business Premium subscribers began the same day. A new Dot has to be created in the ChatGPT desktop app or a desktop browser, not on a phone. The first one comes with extended usage limits during its first month, and chatting with it does not count against ChatGPT’s caps, although the tasks it runs in Codex or ChatGPT Work do. Pro plans cost $100, $200 or $500 a month.

Texting a Dot is next on OpenAI’s list, followed by a fixed monthly fee to add more Dots or to make one faster and able to take on more work, with prices still unannounced. The picture OpenAI showed on stage goes further: a small team of Dots for every user, each one working away on its own computer.

Tags: artificial intelligence, OpenAI, AI agents, chatgpt, GPT-6 Astra, Meta Muse