OpenAI has simplified its API usage-tier system from five paid levels to three, making it easier for developers to qualify for higher rate limits. The new structure consists of Build, Launch and Grow, with the highest Grow tier now requiring $500 in cumulative API credit purchases instead of the previous $1,000 threshold.
The change, announced on October 6, 2026, is primarily about access to higher API capacity rather than a reduction in the price of OpenAI’s models. Existing paid organizations are automatically moved to the corresponding new tier, meaning developers do not need to make an additional payment to migrate. OpenAI says the new structure is designed to make higher rate limits easier to reach.
Source and verification note (7 October 2026): OpenAI’s dated 6 October announcement, API changelog and current tier guidance verify the three paid tiers, cumulative purchase thresholds and automatic migration. Independent analyses by CellCog and ThreatFrontier compare archived and current tables, while JLS separately examines the developer implications. The $500 figure is a cumulative API credit-purchase threshold for Grow, not a monthly subscription or model-token price.
Key takeaways
- OpenAI has reduced its five paid API tiers to three.
- The new paid tiers are Build, Launch and Grow.
- Grow, the highest tier, now requires $500 in cumulative API payments, down from $1,000.
- Build requires $5 in cumulative credit purchases, while Launch requires $100.
- Grow retains a $200,000 monthly usage limit.
- Existing paid organizations are migrated automatically.
- The change affects usage tiers and rate limits, not the underlying per-token prices of OpenAI models.
- Higher tiers provide substantially greater requests-per-minute and tokens-per-minute capacity.
- The change could particularly benefit developers building high-volume AI agents and applications.
OpenAI has simplified the API ladder
OpenAI’s previous API system had five paid usage tiers.
That structure required developers to progress through multiple levels before reaching the highest available capacity.
The company has now compressed that system into three paid tiers:
| New tier | Cumulative API credit purchases | Monthly usage limit |
|---|---|---|
| Build | $5 | $500 |
| Launch | $100 | $5,000 |
| Grow | $500 | $200,000 |
OpenAI’s current documentation also lists a Free tier for eligible users, with a $100 monthly usage limit. The three-tier figure therefore refers specifically to the paid usage tiers, not the entire system.
The biggest change is at the top.
Previously, developers needed $1,000 in cumulative API payments to qualify for the highest tier. That threshold has now been cut in half to $500.
The $500 figure is not a new API price
One important clarification is necessary because the headline can be misleading.
OpenAI has not reduced the cost of its API to $500.
The $500 figure is the amount of cumulative API credit purchases required to qualify for the Grow usage tier.
It is effectively an access threshold.
Developers still pay the applicable model prices for the tokens and services they consume. The tier determines the amount of traffic their organization can handle before hitting rate limits and the approved monthly usage ceiling.
OpenAI’s Help Center explicitly distinguishes the usage-tier qualification system from API pricing and configured spend limits.
That means a developer does not receive $200,000 of free API usage after paying $500.
Instead, reaching Grow makes the organization eligible for a much higher approved monthly usage limit, subject to actual API charges.
Why rate limits matter
API rate limits determine how quickly an application can use an AI model.
The two most important measures are:
- RPM: requests per minute
- TPM: tokens per minute
An application can hit either limit.
For example, a developer could have plenty of token capacity remaining but still be unable to send more requests because the application’s RPM limit has been reached.
Conversely, an application making relatively few but extremely large requests could reach its TPM limit first.
OpenAI says rate limits are applied at the organization and project levels and can vary by model.
For developers operating AI applications at scale, these limits can become a significant infrastructure constraint.
Grow provides a major jump in capacity
The new Grow tier provides substantially more capacity than Build and Launch.
For OpenAI’s Astra, Sol and Terra model families, the standard limits are:
| Tier | Requests per minute | Tokens per minute |
|---|---|---|
| Build | 5,000 | 1 million |
| Launch | 10,000 | 4 million |
| Grow | 15,000 | 40 million |
For Luna, the limits are considerably higher:
| Tier | Requests per minute | Tokens per minute |
|---|---|---|
| Build | 5,000 | 2 million |
| Launch | 10,000 | 10 million |
| Grow | 30,000 | 180 million |
These are standard rate limits; individual models or specialized service tiers can have different limits.
The difference becomes particularly significant for applications handling long prompts or large numbers of concurrent AI-agent tasks.
The change is especially relevant for AI agents
The timing of the change is important.
OpenAI has been expanding its developer platform toward agentic applications, including tools, computer use and long-running workflows.
Agents can make many model calls during a single task.
An agent might:
- Read a user’s request.
- Plan a task.
- Search for information.
- Call a tool.
- Inspect the result.
- Reason about what to do next.
- Call another tool.
- Generate a final answer.
A single user request can therefore generate substantially more API traffic than a traditional chatbot interaction.
Higher TPM and RPM limits become increasingly important as developers build these systems.
OpenAI’s new Grow tier gives developers access to up to 180 million tokens per minute on Luna under the standard limits, creating substantially more headroom for high-volume workloads.
Developers now reach the top tier with half the previous threshold
The most direct benefit is for organizations that previously needed to spend $1,000 to reach the highest API usage tier.
Under the new structure, the threshold is $500.
OpenAI’s developer community announcement described the change as making it easier to qualify for higher API rate limits and confirmed that the old $1,000 threshold has been reduced to $500.
That represents a 50% reduction in the cumulative payment requirement for the top tier.
It could make high-capacity API access more accessible to smaller startups and independent developers that are already generating meaningful API workloads but do not want to commit $1,000 merely to unlock higher limits.
Existing customers do not have to do anything
Another important part of the announcement is automatic migration.
OpenAI says organizations already on paid tiers will move to the new tier structure automatically.
There is no additional payment requirement simply because the tier names have changed.
OpenAI’s Help Center says organizations previously in Tiers 1 and 2 move to Build, Tiers 3 and 4 move to Launch, and Tier 5 moves to Grow.
This prevents the restructuring from creating unnecessary friction for existing customers.
The migration also means that the change is not simply a new pricing plan being offered to new users. It changes how the existing API customer base is classified.
What happened to the old five-tier system?
Before the change, OpenAI’s paid API system consisted of five tiers.
The previous structure progressively increased both monthly usage allowances and the amount of cumulative payment required to qualify for each level.
The new system effectively removes two intermediate stages.
Instead of gradually moving through five paid levels, developers now have three:
Build → Launch → Grow
That is a much simpler progression.
For OpenAI, the benefit is potentially easier communication and administration.
For developers, it makes it easier to understand what they need to do to reach a particular level of capacity.
The middle tier is also more generous
The changes are not limited to Grow.
Launch now requires $100 in cumulative API purchases and provides a $5,000 monthly usage limit.
That is a significant increase over the previous intermediate structure.
Independent analysis of OpenAI’s old and new limits indicates that each new tier provides the same or greater monthly usage capacity compared with the corresponding previous levels.
This matters for startups that are beyond experimentation but are not yet operating at the scale of a major AI application.
They can reach Launch without having to progress through several small increases.
OpenAI still maintains separate spend controls
The usage tier should not be confused with an organization’s own spending controls.
OpenAI allows organizations and projects to configure spend alerts and hard spend limits.
A spend alert notifies administrators when spending reaches a chosen level but allows API traffic to continue.
A hard spend limit can cause affected API requests to return a 429 error after the configured threshold is reached.
This means a company can have Grow-level API capacity while still setting its own lower internal spending ceiling.
That distinction is important for businesses building applications that could generate unpredictable amounts of automated API traffic.
Higher rate limits do not mean unlimited access
The new Grow tier is significantly more generous, but it does not remove rate limiting.
An organization can still encounter:
- Requests-per-minute limits
- Tokens-per-minute limits
- Monthly usage limits
- Project-level limits
- Spend limits
- Temporary overload conditions
OpenAI also distinguishes between traffic that increases too quickly and temporary model overload.
A rapid increase can result in a 429 response with a slow_down code, while temporary model overload can produce a 503 response with a server_is_overloaded code.
That distinction is particularly important for developers running large-scale applications.
Simply upgrading to Grow does not mean an application can suddenly send unlimited traffic.
Why this matters for startups
For startups, API capacity can become a bottleneck surprisingly quickly.
An early-stage company might begin with a few hundred users and relatively modest API traffic.
If its application becomes popular, usage can increase dramatically.
AI agents can accelerate that growth because one user action can trigger many model calls.
A higher rate-limit tier can therefore become part of the infrastructure required to scale.
The reduction in the Grow qualification threshold could give startups a faster route to that capacity.
It also reduces the amount of capital they need to put through the API before reaching the highest standard tier.
The change fits OpenAI’s broader developer strategy
OpenAI’s API platform has been expanding rapidly.
On October 6 alone, the company announced the Decisions API in beta, which allows developers to use GPT-6 Luna to make structured decisions for routing models, tools and actions. OpenAI says the Decisions API can make those decisions up to 10 times faster than using GPT-6 Luna through the Responses API.
That release and the rate-limit restructuring point in the same direction.
OpenAI is trying to make its API platform easier to use for developers building increasingly complex applications.
The company is not simply selling access to a chatbot model.
It is building infrastructure around agents, routing, tools and high-volume application workloads.
What developers should do now
Existing paid customers do not need to manually migrate.
However, developers operating close to their current limits should check their organization’s Limits page.
OpenAI recommends monitoring the relevant RPM and TPM limits rather than assuming that monthly usage capacity alone determines how much traffic an application can process.
Developers should also maintain appropriate retry and backoff logic.
OpenAI recommends respecting Retry-After headers and using exponential backoff with jitter for temporary rate-limit errors.
This becomes even more important when applications automatically generate traffic through agents.
The bigger picture
OpenAI’s tier restructuring looks like a small administrative change, but it reflects a larger shift in the API business.
As AI applications move from simple prompt-and-response interactions toward autonomous agents and complex workflows, developers need more predictable access to model capacity.
Reducing five paid tiers to three removes friction, while cutting the Grow qualification threshold from $1,000 to $500 makes the highest standard tier accessible sooner.
The important number is therefore not the “$500 cost” of the top tier. It is the $500 cumulative-payment threshold required to reach Grow, which comes with significantly higher API capacity.
Looking Ahead
The next question is whether OpenAI will continue lowering barriers to high-volume API access as competition among AI platforms intensifies. Startups increasingly have alternatives from companies such as Anthropic, Google and open-model providers, making developer experience and infrastructure economics an increasingly important part of model competition.
For developers, the practical impact is straightforward: fewer usage tiers, automatic migration and a much lower threshold for the highest standard tier. As agentic applications consume more tokens and make more API calls, the value of higher rate limits could become increasingly important—even as developers continue to manage their own spending controls.
For related OpenAI developer updates, see our coverage of the Decisions API public beta and free Codex Auto-review checks. Their eligibility and billing terms are distinct from these API usage tiers.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
