ChatGPT's bigger tasks will eat your paid allowance faster from December

The short version

• From December, most paid ChatGPT users will share one usage allowance across ChatGPT and OpenAI's coding tool, Codex.
• Harder tasks and stronger models will use that allowance faster, while Instant chat won't use it at all.
• If your company pays for ChatGPT Business, set spending limits now, because by default no limits are in place.

What is changing

OpenAI has updated its help pages with a short note on how paid ChatGPT plans will work from December. Its Work mode can carry out longer tasks by itself. OpenAI is moving those features into the main ChatGPT experience.

As part of that move, usage on Plus, Pro and Business plans will be counted together across ChatGPT and Codex. OpenAI will measure it by the amount of text a model reads and writes. How much of your allowance a job uses depends on the model you pick and the task you give it.

So a quick question and a long research job will no longer feel the same to your budget. Instant chat, the fast everyday mode, won't draw from the allowance.

Users will see their remaining allowance and reset times under Settings, then Usage. When someone hits the limit, they have 3 choices: wait for it to reset, move to a higher plan, or buy credits.

OpenAI says it expects most users won't have to wait for a reset more often than they do now. The note doesn't say how large the new allowances will be.

What is new for companies

ChatGPT Business already has the machinery for this. Each seat comes with included usage for advanced features such as Codex. When that runs out, a person can keep working by drawing on a pool of credits the workspace has bought.

Buying credits is optional. If the pool is empty, the feature is blocked until a workspace owner adds more. Basic features such as search and file upload don't need credits.

A finance manager should look at 3 details:

• Limits are off by default. All seats and users start with no monthly credit limit. Owners and admins have to add one.
• Automatic reload can run without a ceiling. If an owner turns it on and leaves the monthly recharge limit blank, purchases are unlimited each month.
• Credits don't come back. Purchased credits last 12 months. They generally can't be refunded, apart from cases such as billing errors or unauthorized use.

Enterprise and Edu plans work differently, and OpenAI's note on the December change doesn't list them. Those customers buy a shared pool at contract level, and everyone draws from it. Once it's gone, advanced features pause unless the owner has allowed overages or bought more.

What it means for your company

Many companies think of a ChatGPT seat as a simple fee per person. This change moves it closer to an electricity bill, where heavy use shows up.

That is a fair way to charge for AI that works through long tasks. It also changes who needs to pay attention. A team running long tasks through ChatGPT can use far more than a colleague who asks for email drafts. In a company with many seats, a few heavy users may hit their limits while most never notice.

Block extra usage, and your most active teams may stop mid-task. Allow it without limits, and your bill can grow with nobody having approved it. Companies in Indonesia that pay for software in foreign currency will also want a clear monthly figure to plan around.

What to do before December

• Find out who your heavy users are. Workspace owners can download usage reports from the billing settings.
• Set a monthly credit limit for each seat type, then add exceptions for the teams that need more.
• If you use automatic reload, enter a monthly recharge limit. Leaving it blank means there is no cap.
• Turn on usage alerts so that the budget owner hears about high use before the invoice arrives.
• Tell staff that the choice of model now affects a shared budget. Everyday questions can stay on Instant chat.
• On an Enterprise contract, ask your OpenAI account team how the change affects you, and check what your contract says about overages.