ChatGPT Pro usage limits now: what each tier gets before December

OpenAI reversed a cut it had announced only a month earlier. Tibo, the OpenAI staffer who fronts ChatGPT and Codex on X, posted that the company was “doubling all usage limits” for the Pro 100, Pro 200 and Pro 500 tiers and Business Premium, starting that day. He added that OpenAI had listened to subscribers and would leave the Pro 200 allowance alone.

The doubled limits mostly apply to two features that already share one pool of usage on eligible plans:

• Codex, OpenAI’s coding agent, which writes, reviews and ships code for you.
• ChatGPT Work, the mode for longer jobs such as research reports, documents and presentations.

Plus subscribers on the $20 plan get nothing new, since OpenAI’s help pages say Plus usage is unchanged. Nobody’s monthly bill went up.

To see why Pro subscribers were angry, go back to DevDay on September 29. OpenAI launched the Pro 500 plan there and reopened Pro 200 to new sign-ups, but with less included usage. Under that plan, Pro 200’s Work and Codex allowance would have been halved, to 10 times what Plus gets instead of 20 times. The weekly cap on GPT-6 Pro chat messages would have gone to 100, down from 200.

Existing subscribers were told they could keep their old limits until October 29, and would get a one-off bundle of usage credits worth $2,500, good until the end of the year, to soften the blow.

That plan is now dead. OpenAI’s help page on Pro tiers says the reduction from its September 29 email is canceled. People already on the higher Pro 200 allowance keep it, and anyone who had been put on the lower allowance has been doubled up to match. The lower weekly GPT-6 Pro message cap on Pro 200 has also been raised to the higher figure.

Because OpenAI doesn’t publish exact limits, the honest comparison is relative. Here’s how the plans line up after the change:

• Pro 100 ($100 a month): the Work and Codex allowance is now twice what it was. It’s the natural step up for heavy Codex users who keep running dry on Plus.
• Pro 200 ($200 a month): no cut. Everyone on this tier now sits at the higher allowance, which is bigger than Pro 100’s.
• Pro 500 ($500 a month): also doubled, and still the largest allowance of the three. It’s the only Pro plan with Ultrafast, which OpenAI says can produce as many as 300 tokens per second in Codex, roughly eight times standard speed.
• Business Premium: the shared allowance doubles as well.

Every Pro tier also now comes with a monthly API credit that matches its price: $100, $200 or $500 to spend on OpenAI’s developer platform. To claim it, you link the subscription to an API organization, and it must have been active for at least 7 days. Unused credit vanishes when each billing cycle ends, so it doesn’t build up.

The changes were announced in US dollars. In the UK and Canada, check the checkout page for your local price, which depends on tax and exchange rates.

A bigger limit is not the same as twice as many finished tasks. OpenAI’s own help page makes that point: what you actually get done depends on how heavy each job is. If you run the same sort of work every week, the doubling simply gives you more breathing room. If you switch to heavier models or settings, a lot of that extra headroom disappears.

Some choices burn through usage much faster. OpenAI’s pricing pages say Fast mode in Codex eats the allowance 2.5 times faster on supported models, and image generation also drains it several times faster than normal, depending on size and quality.

Two things stayed the same. Dots, OpenAI’s always-on assistants, keep their own extended limits for now, separate from your default usage. Dots haven’t launched for Pro users in the UK, Switzerland or the European Economic Area, though. And OpenAI has said it won’t bring back the five-hour limit on Pro 200, so heavy users don’t face a short rolling cap there.

The next shift is bigger than the doubling. In an email to subscribers, OpenAI said ChatGPT and Codex will soon draw from a single allowance. Its help documentation says this starts in December for Plus, Pro and Business, though no exact day has been given.

Today, ordinary chats sit outside the Codex and Work pool. From December:

• Usage is counted in tokens, the small chunks of text that AI models read and write.
• Instant chats are included in your plan and won’t count against the shared limit.
• Advanced models and complex tasks take a bigger bite.
• You’ll see one combined figure, plus reset times, under Settings and then Usage.

There’s a catch for Instant fans. ChatGPT can switch on reasoning by itself even when Instant is selected, and the help pages don’t yet say how that will be counted. If you want to stay in control, look at the Higher intelligence option under Settings and then General.

OpenAI is also testing a merged Chat and Work experience with a small group of users over the coming weeks, so you can go from a normal conversation to a bigger task without switching modes. Codex will stay a separate experience in the ChatGPT desktop app.

Start with how you use Codex and Work, not with the price tag.

1. Check your usage first. Open the usage page in ChatGPT’s settings, or type /status inside the Codex command-line tool, and note where you stand every week or two.
2. Hitting the cap on Plus? Pro 100 is the obvious next step, and with its allowance doubled it now covers far more than it did a week ago.
3. Regularly running out on Pro 100? Pro 200 gives you the larger allowance, and the threatened cut is off the table.
4. Need raw speed? Only Pro 500 has Ultrafast. If you can live without it, Pro 200 is the better value.
5. Planning to cancel over the cut? That reason is gone. You don’t have to cancel and resubscribe to get the new limits, because active subscriptions update on their own.

Before December, it also pays to trim waste. Be precise in your instructions, attach only the files that matter, and switch off MCP servers (links to outside tools) you aren’t using, since every one of them pads out each message and quietly uses up more of your limit.