ChatGPT Free Is Unlimited Now: 5 Checks Before You Cancel Plus

The $240 a year leaving your card for ChatGPT Plus was mostly buying room to keep typing. On August 6 OpenAI gave that room away, and it gave away the typing part only.

Both halves of that sentence come from the same announcement. OpenAI wrote that free and Go users get "unlimited text chats and access to a new Think button," and one line later, "Limits will still apply for file uploads, images and other tools." Most of the coverage carried the first sentence. The second one is the one that decides whether your subscription is now dead weight.

There is a freshness problem worth naming before you read anyone else's version of this. A large share of the "is ChatGPT Plus worth it" pages still sitting near the top of search were written before August 6 and still describe the free tier as ten messages every five hours. That number is gone. Any comparison built on it is now describing a plan that no longer exists.

What actually changed on August 6, and what did not

One model swap and one cap removal, both on the free and Go tiers. OpenAI made GPT-5.6 Luna the default for free and Go users and removed the limit on text conversations.

The model swap matters more than it sounds. OpenAI reported that "responses containing at least one factual error were about 62% less common with GPT-5.6 Luna and 68% less common with GPT-5.6 Sol than with GPT-5.5 Instant." The free tier did not simply get more turns of a weaker model. It got a different default, plus the Think button.

What did not change is the shape of every other limit. Uploads, images, voice and deep research kept their caps, and OpenAI has published no numbers for any of them on any tier. The free library that holds your stored files is capped at 500 MB, which is the one hard figure in the current plan documentation.

What does "unlimited" cover, and where does it stop?

Text, with a condition written into the sentence. The help center says free users have "unlimited everyday text chats, subject to abuse-prevention safeguards," and it never says where those safeguards begin.

That clause is not decoration. It means the ceiling exists, has not been quantified publicly, and can be reached by someone using the product hard enough. I have not run a free account into it and will not claim a number I have not measured, and neither should any page that tells you the free tier is now infinite.

The ad question deserves its own line, because for some readers it is the entire subscription. OpenAI's ads page describes the ads as running on the free and Go tiers and states that "Pro, Business, Enterprise, and Education tiers will not have ads." Plus is absent from that sentence, which is the one gap I could not close from the page itself. The same page says the ads do not influence the answers you get and that advertisers cannot reach your conversations, history or memory. Rollout is a country-by-country pilot, and the published expansion list names the UK, Mexico, Brazil, Japan and Korea, so your account this month is not a forecast of your account in November.

Which plan fits which user? I made a model do the arithmetic

Three user types, two of which should stay on Plus for reasons that have nothing to do with message counts. I gave Claude Opus 5 a three-line fact list, three user profiles, and an instruction to write "unknown" rather than invent any limit it was not given.

The method caveat matters here. The model was working from the four published facts above and nothing else, so this is a test of plan arithmetic and of what a fact list cannot answer. It is not a review of the product, and the model had no ability to check a real account.

Look at the first line under Plus to Free. The model wrote $12, caught itself inside the same line, and rewrote it as $20. I left the run as it came out, including that. The final figures are right, and I checked every one of them in a separate arithmetic pass before publishing this, which is the part of the workflow that keeps a confident wrong number from reaching you. A model that corrects itself mid-sentence is doing better than a model that does not, and neither is a reason to skip the recheck.

The Go row is the interesting one. It looks like the sensible middle, and it is the one choice the published material cannot justify, because nothing tells you whether Go's caps sit above free's caps. You would be paying $96 a year for a difference nobody has quantified.

Five checks before you cancel

Run these in order. The first two settle most cases, and the fifth is the only one that produces evidence rather than an opinion.

1. Count last month, do not estimate it

Open your ChatGPT history and count two things: how many files you uploaded, and how many images you generated. Not your sense of it. The count. Everything the paid tier still protects sits in those two numbers, and people who are sure they "barely use it" often find four PDFs and a slide deck in there. People who are sure they need it sometimes find thirty days of text.

2. Name the paid-only feature you would lose

A limit and a gate are different problems. Unlimited text does nothing about a feature the free tier cannot open at all, and the clearest current example is account linking: ChatGPT's Finances view, which connects real bank and brokerage accounts through Plaid, reached US Plus and Pro users on June 25, 2026. If you can name a gate like that and you use it monthly, the decision is already made. If you cannot name one, you are paying for headroom.

3. Put a number on the ad question

Ads run on free and Go. If seeing them would end your relationship with the tool, then your subscription is not buying limits at all, it is buying quiet, and the published tier list is the thing to check before you assume Plus stays quiet. Decide what quiet is worth to you per month before you look at the price, then compare. Most people who do this in the other order talk themselves into the number they were already paying.

4. Treat the $8 middle tier as the trap it currently is

Going Plus to Go saves $144 a year and delivers an unknown. OpenAI publishes no caps for any tier, so the upgrade you would be buying over free is undocumented, and the ads come along regardless. Between a documented $0 and an undocumented $96, the $0 is the one you can reason about.

5. Spend one billing cycle behaving like a free user

Keep paying, and use the product for thirty days as if you had already canceled: no uploads, no image generation, no deep research. If nothing breaks, cancel with evidence instead of a hunch. If something breaks on day four, you just learned what your $240 is for, and it cost you nothing to find out. This is the same trick that works on the forgotten subscriptions sitting in your bank statement, and it is the only one of these five checks that produces a fact.

How to cancel, and what happens to the days you already paid for

Cancel where you bought it, not where you use it. If the charge appears on an Apple or Google receipt rather than an OpenAI one, the subscription lives in that store's subscription settings, and turning it off inside ChatGPT will not stop the billing; if you subscribed on the web, it is in the ChatGPT settings under your plan. Pull up the receipt from your last charge before you go looking, since that email names the biller and settles the question in about ten seconds. Then confirm the exact path in OpenAI's help center rather than trusting a screenshot from a blog post written two interface revisions ago, and check for the cancellation confirmation afterward, because an unconfirmed cancel is the single most common way a subscription survives being canceled. While you are in there, an audit of the other recurring charges on the same card usually pays better than this one decision does.

The rule that outlives the model name

Luna, Sol, whatever lands in November. The model names in this article have a shelf life of about a quarter, and the decision rule underneath them does not move: you are not choosing a model, you are choosing which caps you bump into.

That is why the plan question keeps coming back every month instead of being settled by any single announcement. Text was capped, so people paid. Text is uncapped, so some of those people should stop paying. When uploads or images or whatever comes next gets uncapped on the free tier, the same two-number check in section one gives you the answer again without anybody rewriting a comparison table. It is the rule I keep pinned above each of these announcements, and the one WONDY will still be running here when the next model name lands.

My own read, stated plainly: for a text-only user, $240 a year is now hard to defend, and I would cancel. For anyone whose month contains real uploads, Plus is still the cheaper mistake, because the caps that would replace it have never been published and you would be discovering them at the worst possible moment. And if the money does come free, a payoff order for what you already owe beats letting $240 dissolve back into the month.

FAQ

Now that ChatGPT free has unlimited text chats, what do I actually lose if I cancel Plus?

You lose the caps that never moved. OpenAI's August 6, 2026 announcement says free and Go users get unlimited text chats, and the next sentence in the same post says "Limits will still apply for file uploads, images and other tools." So the thing you give up by canceling is headroom on uploads, image generation, voice and deep research, plus whatever ad-free status your tier carries. You also give up the paid-only features, which are gates rather than limits: account linking for the Finances view rolled out to US Plus and Pro users on June 25, 2026, and no amount of free text chatting reaches it. If your last month contained zero file uploads and zero generated images, the $240 a year is buying you room you are not standing in.

Is ChatGPT free really unlimited?

For everyday text chats, with a condition written into the sentence. The ChatGPT plans and pricing help center says free users have "unlimited everyday text chats, subject to abuse-prevention safeguards," and it does not publish the threshold where those safeguards start. The pricing page lists the free tier as "Unlimited text chats with GPT-5.6 Luna." Everything outside text kept its cap: file uploads, images, voice and other tools, plus a 500 MB library for stored files. Treat unlimited as a description of one activity rather than a description of the plan, because no number has been published for any of it.

Does the $8 Go plan get rid of the ads?

No. OpenAI's ads page describes the ads as running on the free and Go tiers, and states that "Pro, Business, Enterprise, and Education tiers will not have ads." Plus is not named in that sentence, which is worth checking against the live page if ad-free status is the reason you keep paying. The rollout is a country-by-country pilot, and the published expansion list names the UK, Mexico, Brazil, Japan and Korea, so what your account shows this month is not a guide to what it shows next quarter. OpenAI also states the ads do not influence the answers you get and that advertisers cannot access your conversations, history or memory.

How much do I save by canceling ChatGPT Plus?

$240 a year going to free, or $144 a year going to the Go plan. Plus is $20 a month, so $20 x 12 = $240. Go is $8 a month, so $8 x 12 = $96, and $240 - $96 = $144. The Go route is the one that looks smarter and is harder to justify: OpenAI has not published the numeric limits on any tier, so you cannot tell from the published material whether Go raises the upload and image caps above free at all. That trade is a known $144 against an unknown ceiling. Prices read at openai.com/chatgpt/pricing on August 13, 2026, in US dollars.

The same "what am I actually using" question applies well beyond one subscription. If this one ends in a cancellation, the Korean-market version of this decision covers the same plan boundary for readers billed in won, free and open-source money tools cover a surprising amount of what people pay AI subscriptions to do, and the self-audit on which parts of your own job AI is closing in on is the larger version of the same arithmetic.

Sources and limits of this article

• OpenAI, "Improving GPT-5.6 Sol in ChatGPT," August 6, 2026 (GPT-5.6 Luna as the default for free and Go; "unlimited text chats and access to a new Think button"; "Limits will still apply for file uploads, images and other tools"; factual errors about 62% and 68% less common than GPT-5.5 Instant): https://openai.com/index/improving-gpt-5-6-sol-in-chatgpt/
• OpenAI Help Center, "ChatGPT plans and pricing" ("unlimited everyday text chats, subject to abuse-prevention safeguards"; 500 MB free library), read August 13, 2026: https://help.openai.com/en/articles/9275245-chatgpt-plans-and-pricing
• OpenAI, ChatGPT pricing page (Free $0, Go $8, Plus $20, Pro from $100 per month; "Unlimited text chats with GPT-5.6 Luna"), read August 13, 2026: https://openai.com/chatgpt/pricing/
• OpenAI, "Testing ads in ChatGPT" (ads on the free and Go tiers; "Pro, Business, Enterprise, and Education tiers will not have ads"; ads do not influence answers; advertisers cannot access conversations, history or memory; expansion list naming the UK, Mexico, Brazil, Japan and Korea), page updated August 11, 2026: https://openai.com/index/testing-ads-in-chatgpt/

Three limits on what you just read. I did not measure the abuse-prevention ceiling on a free account, so nothing here tells you where unlimited stops in practice. The ad experience is a pilot that varies by country and by account, and the published tier language does not name Plus either way, so verify the current page if that is your deciding factor. And all prices are US dollar list prices read on August 13, 2026, before tax and before any app store markup, which means your receipt may not match this page. Plan terms in this market have changed twice in the past year, and the check in section one is built to survive the next change.