ChatGPT as a Mac app: what the free tier covers

Searching ChatGPT as a Mac app turns up two answers that look contradictory. One says it is free. The other says it needs a subscription. Both are accurate, because two different bills are being described: the cost of putting an icon in the Dock, and the cost of the account running behind it. Separating them takes about five minutes and usually saves a monthly charge that was never needed.

Every route to a Dock icon starts at zero

The official desktop app from OpenAI costs nothing to download and nothing to install. It runs on a free account. The requirement that actually gates it is hardware and operating system, not payment: macOS 14 or later, with either Apple Silicon, meaning M1 or better, or an Intel processor. The download page detects which processor the Mac has and serves the matching installer.

The browser route also costs nothing. Safari has an Add to Dock command that creates an icon in the Dock and in Launchpad. Chrome has Install page as app, under Cast, save, and share in the menu, with an install control appearing in the address bar on some sites. No account upgrade is involved in either.

Purpose built tools for turning a website into a standalone Mac app are paid products, but both of the well known ones let you build a first app at no cost. Coherence X6 does, and so does Unite Pro. That means the experiment of turning one site into an app can be run to completion, on any route, without spending anything.

So the honest version of the free answer is this: getting ChatGPT out of the tab strip and into the Dock is free on all three roads. Money enters the picture for two reasons only, and both are worth understanding before reaching for a card.

What the free plan actually covers

The pricing page lists the free tier as text chat without a message cap, with a footnote attached.

Unlimited text chats subject to abuse guardrails. Source: openai.com

Around that, the limits are specific rather than vague. Uploads are limited. Image generation is limited and slower. Voice chats are limited. Deep research is limited. Memory and context are limited. Codex access is limited. Access to the document and spreadsheet side of the product on the desktop is limited as well.

Notice what is not on that list. Nothing about the app itself. The launcher bar that opens on Option Space, the menu bar icon, attaching a file or a screenshot from that bar, the Dock icon: none of those are plan features. They come with the application. That is the single most useful fact for anyone trying to decide whether a subscription is required to get the Mac experience. It is not. The subscription changes how much you can do once you are inside, not whether the window exists.

For a large number of people this settles the question. If the daily pattern is typing questions and reading answers, the free tier plus a free install is the whole answer, and the monthly cost stays at zero.

The ladder, and what each rung buys

When the free limits do start to bite, the ladder is four rungs. The prices below are the ones the pricing page shows in yen.

Plan	Monthly	What it adds
Free	0	Text chat without a cap, everything else limited
Go	1,400	More messages with tools, more uploads, more image creation, longer memory
Plus	3,000	Advanced reasoning models, expanded deep research, projects and scheduled tasks
Pro	from 16,800	Roughly five times the usage, maximum research and memory

Two footnotes on that table are easy to miss. The Go plan carries a note that it may include ads. And the desktop document workspace is listed as available on Plus, Pro, and the business plans, with the free tier marked as limited. If the reason for wanting a Mac app is producing documents and spreadsheets from a desktop window, that is the rung where the feature actually opens.

The gap between rungs is worth a moment of arithmetic. The step from free to Go is 1,400 yen a month, which is 16,800 yen over a year. The step to Plus is 3,000 a month, or 36,000 over a year. Those are real numbers to weigh against how often a limit gets hit.

Check availability before budgeting for the top plan

Price is not the only thing that can change on a plan page. Availability can too. At the time of writing, the help center carries a banner stating that new sign ups and upgrades to the 200 dollar Pro plan are temporarily paused, and that existing Pro subscriptions continue to renew as usual.

This is the sort of detail that invalidates a plan quietly. Budgeting around the top tier only works if the top tier can be purchased on the day the decision is made. Anyone building a case for the highest rung should confirm that sign ups have reopened rather than assuming the published price implies an open door.

Paying for a plan and then running the wrong app

There is a way to pay and not receive, and it appeared in July. The desktop app was reorganised on July 9th to combine Chat, Work, and Codex, and the previous macOS application was renamed ChatGPT Classic. Classic was not retired. It still receives model updates, bug fixes, security patches, and support for its existing enterprise capabilities, so nothing forces a move. The wording in the migration notes is careful on one point: some new agent features are available only in the new app.

Both applications are free, so this changes nothing about the download. It changes something about the subscription. A plan bought for capabilities that live in the new app produces nothing if the icon being clicked every morning is the old one. The two can sit in the same Dock, with similar names and similar icons, which is exactly the condition under which people click the wrong one for weeks.

The check takes seconds. Open the app, look at the top left, and confirm that ChatGPT and Codex can be selected there and that a Chat and Work toggle appears at the top of the page. If that navigation is missing, the running application is Classic. Whether that matters depends entirely on what the plan was bought for.

Deciding with a week of evidence instead of a feeling

Upgrading on a hunch is how unused subscriptions happen. A week of counting replaces the hunch cheaply.

Count three things. How many times a response was blocked by a limit, such as an upload refused or an image not produced. How many times slow responses at busy hours stopped work. How many times another paid service was opened because this one would not finish the job. Those three numbers cover almost every real reason to move up a rung.

If each of them stays at a handful for the week, the free tier is doing its job and the correct monthly spend is zero. If one of them repeats daily, start at the 1,400 yen rung rather than the 3,000 one, and move again only if the same limit reappears. Climbing one step at a time costs a month of patience and routinely saves the difference between rungs for the rest of the year.

The bill follows accounts, not icons

Here is the point where people overestimate the cost of the Dock approach. Building five app icons does not multiply a subscription by five. The icons are containers. Billing attaches to the account signed into them. Five containers holding one account produce one bill.

The multiplier is the account itself. Two separate accounts, both upgraded, is what doubles the monthly charge. That distinction opens a cheap arrangement that many people miss: upgrade the account that earns money, leave the other on the free tier, and give each one its own container and its own icon. The switching friction disappears at a cost of zero extra yen per month.

Where a wrapper tool starts charging

The second place money can enter is the tool that builds containers. The first app is free on the tools mentioned above. The charge starts when the second one is built, and the shape of that charge differs by product: a one time purchase against a recurring one, with different rules about how many apps can exist.

Counting first makes the comparison short. Write down the sites currently living as permanent tabs, then mark the ones you actively hunt for. Most people land somewhere between four and six: mail, a calendar, a chat tool, an AI assistant, and one or two internal dashboards. Divide the one time price by that number and the per app cost becomes obvious, along with how many months of a subscription it matches.

Operating system floors belong in the same calculation, because they can remove an option before price does. Coherence X6 asks for macOS 13.5 or later. Unite Pro asks for macOS 15 or later. On an older Mac the cheapest viable route may be the only viable route. The Pricing page is where those conditions are stated for this kind of tool, and reading it after counting apps rather than before turns a vague comparison into a single number.

If the count came out at one, stop. One app does not justify buying anything. The browser command handles it, and the decision can be revisited later at no loss, since nothing was spent.

The cost that never appears on a receipt

The last cost is time, and it is spent on rebuilding. Setting an icon, signing in, granting notification permission, and deciding where links open takes a few minutes per app. Doing it twice because the approach changed halfway is the most common waste in this whole exercise.

Three decisions cause almost all of the rework. Which account each container holds. Whether links open inside the app or get handed to a browser. Whether the icons can be told apart in a crowded Dock at a glance. Settle those before building the second app and the rest go quickly. The Guide walks through the build screens in order, which is the fastest way to see which of those decisions appear where, and the Supported services list shows which other daily tabs are candidates for the same treatment once the pattern is working.

What to change first

Install the official app if the Mac is on macOS 14 or later, stay on the free plan, and use it for a week while counting the times a limit actually stopped you. If that count stays low, the correct spend is zero, and the only remaining work is building containers for any second account with Kagemusha.