OpenAI's API now has three paid usage tiers, Build, Launch and Grow. Grow needs $500 in total purchases, down from $1,000, and keeps its $200,000 monthly limit.
OpenAI has replaced the five paid usage tiers of its API with three, named Build, Launch and Grow. The top tier, Grow, now needs $500 in total credit purchases, down from $1,000 for the old Tier 5, and carries the same $200,000 monthly usage limit. Organizations on a paid tier move to their new tier automatically; the Free tier is unchanged. The change was announced in an OpenAI developer community post on 6 October 2026, 18:41 UTC.
What changed
The post, by forum moderator VeitB writing as OpenAI, says five paid tiers become three and that existing paid-tier organizations move over automatically with no action needed. OpenAI's rate-limits guide now lists the tiers below. The docs page carries no date, so the community post is the only dated source. The post's image shows the same per-model limits table as the docs. The post does not say which old tier maps to which new one.
| Tier | Qualification | Monthly usage limit |
|---|---|---|
| Free | Allowed geography | $100 |
| Build | $5 in total credit purchases | $500 |
| Launch | $100 in total credit purchases | $5,000 |
| Grow | $500 in total credit purchases | $200,000 |
Old versus new
An archived copy of the same guide from 22 September 2026, before the change, listed these tiers:
| Old tier (22 Sep 2026) | Qualification | Monthly usage limit |
|---|---|---|
| Tier 1 | $5 paid | $100 |
| Tier 2 | $50 paid | $500 |
| Tier 3 | $100 paid | $1,000 |
| Tier 4 | $250 paid | $5,000 |
| Tier 5 | $1,000 paid | $200,000 |
Old tiers: 22 Sep 2026 archive snapshot; new tiers from OpenAI's rate-limits docs.
Compared with that copy:
- A $5 purchase now gives a $500 monthly limit (Tier 1 gave $100).
- $100 in purchases now gives $5,000 a month (it gave $1,000 as Tier 3).
- The $5,000 limit used to need $250 (Tier 4) and now needs $100 (Launch); the $500 limit used to need $50 (Tier 2) and now needs $5 (Build).
- The $200,000 monthly limit now starts at $500 in purchases, half the old $1,000.
- Waiting periods were already gone before this change. A February 2026 copy of the old page required 7, 14 or 30 days since first payment for Tiers 2 to 5; the 22 September copy lists none.
Rate limits by tier
OpenAI's standard limits table shows requests and tokens per minute by model family:
| Tier | Model | RPM | TPM |
|---|---|---|---|
| Build | Astra, Sol, Terra | 5,000 | 1,000,000 |
| Build | Luna | 5,000 | 2,000,000 |
| Launch | Astra, Sol, Terra | 10,000 | 4,000,000 |
| Launch | Luna | 10,000 | 10,000,000 |
| Grow | Astra, Sol, Terra | 15,000 | 40,000,000 |
| Grow | Luna | 30,000 | 180,000,000 |
These standard limits differ from OpenAI's separate Ultrafast rate limits. (Ultrafast is also sold inside ChatGPT as part of Pro 500; see our Pro 500 and Codex coverage.) (Model pricing is in our GPT-6 launch coverage.)
What it means for developers
Tier upgrades are automatic once total credit purchases reach a threshold, and the console also has an Upgrade tier button under Settings, Organization, Limits. The cost to reach top limits is $500 in cumulative purchases. The monthly usage limit is set by OpenAI for each organization and is separate from the spend limits you configure yourself. Spend alerts only notify, while a hard spend limit makes affected requests return HTTP 429.
Reaching Grow does not remove throttling by ramp rate. OpenAI says a slow_down 429 can occur within RPM and TPM limits when traffic climbs too fast, and suggests raising traffic by no more than 50% every 15 minutes once you pass 1 million input tokens per minute. Check error handlers as well. This is not part of the tier change: the guide already said, by 22 September, that endpoints which returned 503 with slow_down now return 429 with slow_down, while model overload stays 503 with server_is_overloaded.
What to do
- Open Settings, Organization, Limits to see your tier and per-model limits.
- If you sit near a cap, a top-up to the next threshold is enough to upgrade.
- Handle both 429 and 503 in retry code, since the
slow_downstatus code moved on some endpoints. - Set spend limits in addition to the usage limit. (Codex limits on ChatGPT plans are a separate system; see Codex usage limits by plan.)
What is still unclear
- Which old tier maps to which new tier. The post does not say.
- The effective date. The docs page is undated; the post is dated 6 October 2026, 18:41 UTC.