API Rate Limits

OpenAI API rate limits depend on your usage tier (Free, Build, Launch or Grow), which rises automatically as total credit purchases reach $5, $100 and $500. On the Build tier GPT-6 Astra allows 5,000 requests and 1,000,000 tokens a minute, and going over returns HTTP 429.

OpenAI API rate limits are set by usage tier, and as of Nov 6, 2026 there are four: Free, Build, Launch and Grow. An organization moves up automatically when its total credit purchases reach $5 (Build), $100 (Launch) and $500 (Grow), and each tier caps monthly spending at $100, $500, $5,000 and $200,000. On the Build tier, GPT-6 Astra allows 5,000 requests and 1,000,000 tokens per minute. A request over a limit gets HTTP 429, which the right retry logic handles. Your own limits are on the dashboard at Settings > Organization > Limits. For the API in general, see OpenAI API.

Quick facts

Tiers	Free, Build, Launch, Grow, as of Nov 6, 2026
How to move up	Automatic, by total credit purchases
Measured in	RPM, RPD, TPM, TPD, IPM, audio minutes
Scope	Organization and project, not per user
Error	HTTP 429
Where to see yours	platform.openai.com/settings/organization/limits
Batch	A separate queue limit in input tokens

What are the OpenAI usage tiers?

Tier	To qualify	Monthly usage limit
Free	Be in an 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

Source: OpenAI rate limits guide, read Nov 6, 2026.

Tiers rise on their own as purchases cross each threshold. To check or move up, open the limits page and select Upgrade tier in the Usage Tiers section (rate limits guide). The guide describes no separate form for limits above Grow. It also gives no date for this four-tier scheme and does not mention the older numbered tiers, so guides that talk about "Tier 1" to "Tier 5" describe an earlier system.

Limits per model and tier

Each model has its own limits. For the four flagships, as listed on their model pages on Nov 6, 2026:

Model	Build (RPM / TPM)	Launch (RPM / TPM)	Grow (RPM / TPM)
GPT-6 Astra	5,000 / 1,000,000	10,000 / 4,000,000	15,000 / 40,000,000
GPT-6.1 Sol	5,000 / 1,000,000	10,000 / 4,000,000	15,000 / 40,000,000
GPT-6.1 Terra	5,000 / 1,000,000	10,000 / 4,000,000	15,000 / 40,000,000
GPT-6 Luna	5,000 / 2,000,000	10,000 / 10,000,000	30,000 / 180,000,000
gpt-realtime-2.1	400 / 200,000	10,000 / 4,000,000	20,000 / 15,000,000

Sources: model pages for GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Luna and gpt-realtime-2.1, read Nov 6, 2026. RPM is requests per minute; TPM is tokens per minute.

Other limits worth knowing:

Batch queue: GPT-6 Astra can have 3M tokens queued on Build, 200M on Launch and 15B on Grow; GPT-6 Luna 20M, 1B and 15B (GPT-6 Astra, GPT-6 Luna).
GPT-Live: concurrent sessions, 50 on Build, 300 on Launch, 500 on Grow, and none on Free (gpt-live-1).
Long context: requests over 272K input tokens have separate limits (rate limits guide).
Vector stores: 300 requests per minute per vector store for file ingestion (rate limits guide).
Open weights: gpt-oss-20b shows 0 RPM and 0 TPM on every tier, so it cannot be called through OpenAI in practice (gpt-oss-20b). See gpt-oss.

The Free tier's per-model limits were not listed on the model pages read for this article.

How limits are measured

OpenAI counts requests per minute (RPM) and per day (RPD), tokens per minute (TPM) and per day (TPD), images per minute (IPM), and audio minutes per minute for some streaming audio models. You hit the limit on whichever runs out first (rate limits guide). Limits belong to the organization and to each project in it, not to individual users, and spend limits can be set at either level.

Every response carries headers that show where you stand (rate limits guide):

x-ratelimit-limit-requests, x-ratelimit-remaining-requests, x-ratelimit-reset-requests
x-ratelimit-limit-tokens, x-ratelimit-remaining-tokens, x-ratelimit-reset-tokens
x-ratelimit-limit-project-tokens, x-ratelimit-remaining-project-tokens, x-ratelimit-reset-project-tokens
Retry-After, on a limited response

What a 429 error means and how to fix it

A temporary rate limit returns HTTP 429. When traffic rose too quickly, the error type is rate_limit_error with the code slow_down (rate limits guide). A 429 from Flex processing reads "Resource Unavailable" and is not charged (Flex guide). Quota and billing errors are different: they need action (adding credits, raising a spend limit), and retrying will not help.

OpenAI's advice for handling 429s (rate limits guide):

Retry with exponential backoff and random jitter, treating Retry-After as the minimum wait when it is present.
Cap the retries: both the number of attempts and the total time.
Do not double up: if you write your own retries, turn off the SDK's or count them against your limits.
Do not retry quota, billing or other errors that need action.
Ramp up gently: once traffic reaches 1 million input tokens a minute, raise it by no more than 50% every 15 minutes.
Ask for less: set max_tokens close to the answer size you expect.
Move work that can wait to the Batch API, which has its own queue limit.

How to get higher limits

Buy credits: crossing $5, $100 and $500 in total purchases moves you to Build, Launch and Grow (rate limits guide).
Check the tier on Settings > Organization > Limits and use Upgrade tier if it has not moved.
Split traffic across models: each model has its own limits, and GPT-6 Luna's token limits are much higher than Astra's.
Use Batch for bulk work and caching to cut tokens per request (see Batch API).

For what each request costs, see OpenAI API Pricing.

Frequently asked questions

What are the OpenAI API usage tiers? As of Nov 6, 2026, OpenAI has four usage tiers: Free, Build ($5 in total credit purchases), Launch ($100) and Grow ($500), with monthly usage limits of $100, $500, $5,000 and $200,000.

How do I increase my OpenAI rate limit? Buy more credits so your organization crosses the next tier threshold; tiers upgrade automatically, and the Upgrade tier button sits on the limits page under Settings > Organization > Limits.

What does error 429 mean in the OpenAI API? It means a rate limit was hit. Retry with exponential backoff and jitter, wait at least the Retry-After time, and do not retry quota or billing errors, which need you to add credits or raise a limit.

What is the rate limit for GPT-6 Astra? On the Build tier, GPT-6 Astra allows 5,000 requests and 1,000,000 tokens per minute; on Grow, 15,000 requests and 40,000,000 tokens per minute, as of Nov 6, 2026.

Are OpenAI rate limits per user or per organization? OpenAI sets rate limits per organization and per project, not per user.