OpenAI cuts its API usage tiers from five to three — Build, Launch and Grow

OpenAI has simplified its API usage tiers from five to three — Build, Launch and Grow — in an entry its changelog dates 6 October 2026. A usage tier is the level your organization, your account on OpenAI's platform, is placed at, and the entry ties two things to it: a monthly usage limit, and the rate limit for each model — how much you may ask of a model in a given stretch of time. The entry asks nothing of you, but if your application runs close to its limits, the tier your organization now sits in is worth a look this week.

What changed

Three tiers replace five, and they have names: Build, Launch and Grow. Movement between them is automatic: organizations upgrade as their total credit purchases reach each tier's minimum. Two words in that sentence carry the rule.Total means it is the sum of the credit you have bought that counts, not the size of any one purchase. Automatically means the upgrade is not something you apply for.

The entry does not state the minimums, nor the limits attached to each tier. It sends you to OpenAI's usage tiers page for the monthly usage limits and for how to view the rate limit for each model. The figures are on that page, not in the entry.

What it was before

There were five tiers. The entry gives the count, not the names of the five or how they map onto the three, so which old tier became Build, Launch or Grow is a question the changelog does not answer.

What it means for you

Find out which tier you are in.The entry does not say which of the three an existing organization was placed in. Open the usage tiers page it points to, follow its instructions for viewing the rate limit of each model you call, and compare those limits with the numbers your application was built around. If your code carries limits of its own — a queue that sends so many requests a minute, a batch size chosen to stay under a ceiling — this is the moment to check them against what the page shows now.

Know what moves you up.Upgrades follow your total credit purchases: when they reach the next tier's minimum, your organization moves up on its own. If you need higher limits before a launch, the lever the entry names is the credit you buy, and the minimum you have to reach is on the usage tiers page.

Read your rate limits off the tier.The entry sends you to the tiers page to see the rate limit for each model : the limit you work under is read off where your organization stands. When traffic goes past what is allowed, your code has to handle the refusal. The same changelog recorded on 2 September 2026 that API errors now let an application tell traffic that grows too quickly from a model that is temporarily overloaded , and that traffic which increases too quickly can return a 429 error with theslow_downcode. If the response carries a Retry-After header, wait at least as long as it says before you retry; if it does not, back off exponentially, waiting longer after each failed attempt.

Keep your own ceiling apart from the tier's.The tier's monthly usage limit is not the only monthly ceiling on your account. Since 22 July 2026 the platform has had hard spend limits for organizations and projects : a monthly cap you set, after which affected requests return a 429 error once tracked spend reaches it. If you set one, check that it still sits where you want it beside your tier's limits. And because both a spend cap and traffic that grows too fast answer with 429 , read the error itself, not only its status, before you decide which one stopped you.

The source

OpenAI's API changelog, the entry of 6 October 2026, with the entries of 2 September 2026 and 22 July 2026 on the same page.