Twenty dollars divided by 720 hours comes to 2.8 cents an hour.
That is the sticker price of Dots, OpenAI's always-on personal agent, now that it has moved down from the Pro tier to ChatGPT Plus. It is a clean number. It is also the only clean number in the release.
What standing changes
Dots is proactive and it runs continuously. It keeps working on a task without supervision.
The word carrying the weight is not "proactive." It is "unsupervised." Proactive means it speaks first. Continuous means it is present. Unsupervised means you do not have to be. Read together, the product is not a better chat box. It is a stretch of your time that no longer requires you inside it.
That changes the billing shape. A chat session is metered by the request: open a window, type, wait, close. A standing agent is metered by duration. One of those can be sold as a flat monthly fee. The other has to be capped somewhere.
Two meters, two incentives
A per-request product wants you to ask again. A per-hour product wants you to occupy less of it.
OpenAI now sells both, and that is not inconsistency. It is the meter showing through. Moving Dots from Pro to Plus did not change the meter. It changed who is standing on it.
Four thousand doors
Dots connects to more than 4,000 apps. It can also connect to your computer and act on it.
The first number is about entry, not coverage. A standing agent has to get inside the software you already use, and 4,000 is the width of that door. The second capability covers what the list misses: anything not among the 4,000 can still be reached through the screen.
Interface access is the fallback for API access. Together they move the entry point from the integration layer to the display layer.
The model became a part
Dots used to run on GPT-6 Astra. It can now also run on GPT-6.1 Terra.
Two models, one product name. Users buy Dots, not a model SKU. That decouples price from model: the model can change without a price change, and the price can fall without waiting for a new model.
The same release adds an iOS widget, Shortcuts integration, and support for iMessage and RCS. Those are distribution details, and they point the same way. What is being sold is reach.
The allowance is the switch
OpenAI says conversations with the agent do not count against regular usage for Pro and Business Premium subscribers.
OpenAI has not said whether Plus and Business Standard get the same uncapped allowance. That sentence decides what $20 buys, and it is absent.
An uncapped allowance inside a capped subscription is a subscription inside a subscription. A standing agent's consumption does not resemble chat. Chat usage follows your hands. An agent runs in the background for hours across several apps, and none of that passes through your hands. Meter it separately and $20 is not the whole price. Leave it unmetered and the monthly boundary stops meaning anything.
There are several ways to meter it. Per conversation turn is the easiest to explain and the least tied to cost, since an agent may run all day without replying once. Per hour tracks cost more closely and is invisible to the user. Per tool call is closest to real consumption and hardest to explain to a normal customer. OpenAI has not said which one it uses.
Enterprise got the definite sentence
Every Business Standard user had access on the same day. Plus users got a test, open to a small number of them.
Same feature, same day, two rollouts. The business tier got a date. The consumer tier got a list nobody has seen: who is on it, when your turn arrives, and how long the test runs are all unstated.
Read that as sequencing rather than favoritism. A limited-seat test on the consumer side is how a usage model gets calibrated before it is priced for a large base. The enterprise side can absorb a flat allowance without a public argument about fairness.
No end date
OpenAI has also not said whether $20 users will keep Dots in the long run.
The distinction is narrow and it matters. The question is not whether Dots works on Plus. It is whether it will still be there next year. A test can add users, drop users, swap models and rewrite limits from the product side, without a migration notice or a refund.
So the price cut is real and its size is unknown. $20 a month is 2.8 cents an hour only if all 720 hours count, and the allowance decides whether they do. Three things are confirmed: Plus costs $20, Dots is being tested with a small group, and Pro and Business Premium conversations sit outside the regular allowance. One is a price. One is a scope. One is an exception.
Whether Plus and Business Standard fall under that exception is the open item. Until it closes, the discount is a headline, not a rate.
*Disclaimer: This article is based on publicly available information and is for reference only. It does not constitute investment advice.*
