Rate limits and usage
Plan AI traffic before rate pressure reaches users.
Understand request, token, and image monitoring; handle 429 responses safely; and review thresholds before customer-facing traffic grows.
Last updated: July 6, 2026. Examples use safe operational placeholders and do not include credentials or customer data.
Production rule
Treat limits as product controls. Monitor trend lines and threshold states before launch, then decide which workflows should pause when a hard limit is reached.
What is counted
Usage is tracked across requests, tokens, and images.
Each usage type answers a different capacity question: how often your app calls the API, how much text it processes, and how many image request items it returns.
Request limits
Each API call is monitored for service health, abuse prevention, logs, and rate-limit behavior.
Token limits
Chat and embeddings usage count input and output tokens so teams can track the volume behind each feature.
Image limits
Returned image items count separately from text traffic so visual workflows can be monitored and planned on their own.
Unlimited packages
Choose a package that matches your billing rhythm.
Public packages are Unlimited Weekly and Unlimited Monthly. Usage remains monitored for service health, logs, rate-limit behavior, and launch planning.
| Package | Price | Access | Delivery | Recommended fit |
|---|---|---|---|---|
| Unlimited Weekly | $19 / week | GPT and Codex API access with unlimited token consumption and 4 concurrent connections | 10 minutes to 5 hours | For builders who need OpenAI-compatible GPT-5.5 and Codex API access for a short launch, test, or delivery sprint without a monthly commitment. |
| Unlimited Monthly | $76 / month | GPT and Codex API access with unlimited token consumption and 4 concurrent connections | 10 minutes to 5 hours | The best ongoing value for GPT-5.5 and Codex API work, with a $59 first Monthly payment when the reward code is configured in Stripe. |
Rate behavior
Rate states stay informational until a limit blocks a request.
The rate state helps operators act before access is denied. Use warning and urgent states to tune traffic, then treat a capacity 429 as a package or traffic decision.
Usage is inside the normal operating range.
Keep monitoring requests, tokens, images, rate state, and estimated cost.
Usage is getting close enough to plan ahead.
Review traffic sources, model choices, and upcoming launch volume.
The workspace is close to one or more configured rate limits.
Reduce noncritical traffic, shrink request content, or move to a better-fit package.
The next request may be denied if it would exceed a configured rate limit.
Treat this as a hard stop until traffic pressure drops or capacity changes.
429 handling
Separate short rate pressure from capacity pressure.
A 429 can mean a temporary rate limit or capacity pressure. Read error.code, then choose retry, traffic shaping, or upgrade action.
Retry rate_limit_exceeded with Retry-After when the header is present.
Do not immediately retry request_quota_exceeded, token_quota_exceeded, or image_quota_exceeded.
Reduce noncritical traffic or route background work later when capacity pressure is urgent.
Surface rate state before customer-facing workflows depend on remaining capacity.
{
"error": {
"message": "Quota exceeded for this workspace.",
"type": "rate_limit_error",
"code": "request_quota_exceeded"
}
}async function handleLimitedResponse(response) {
if (response.status !== 429) {
return response;
}
const body = await response.json();
const code = body.error?.code;
if (code === "rate_limit_exceeded") {
const retryAfter = Number(response.headers.get("retry-after") ?? 1);
await new Promise((resolve) => setTimeout(resolve, retryAfter * 1000));
return null;
}
if (code?.includes("quota_exceeded")) {
await recordQuotaPressure("Quota pressure needs a package or traffic decision.");
return response;
}
return response;
}Upgrade path
Make the rate decision before the launch spike.
If usage moves from warning to urgent during normal traffic, review request shape, retry behavior, or the billing rhythm before the next release or customer rollout.
Tune first
Reduce payload size, avoid duplicate calls, cache stable results, and move noncritical work to a queue.
Move to Monthly
Use Monthly when validation work is ongoing and the first-month reward code makes the longer billing rhythm worthwhile.
Request walkthrough
Use a walkthrough when routing policy, support expectations, or procurement requirements need a custom review.
Document limit behavior
Decide which workflows pause, degrade, or show an operator message when quota is exhausted.
Monitoring
Watch usage shape, not just totals.
Rate pressure is easier to control when you know which product flow, model, endpoint, or background job is changing the curve.
Request volume and rate-limit pressure for the current period.
Input, output, and total tokens by endpoint or model.
Image request items by product workflow.
Quota state movement from healthy to warning, urgent, or blocked.
429 responses split by rate_limit_exceeded and quota codes.
Estimated cost trends before a campaign or public launch.
{
"workspace": "workspace_example",
"period": "2026-07",
"usage": {
"requests": 74200,
"tokens": 18400000,
"images": 420
},
"quotaState": "warning"
}FAQ
Rate limit and usage questions.
What counts against quota?
Weekly and Monthly are public API packages with unlimited token consumption and 4 concurrent connections, while requests, tokens, and image request items are still monitored for service health, abuse prevention, logs, and rate-limit behavior.
How should my app handle 429 responses?
Check whether the code is rate_limit_exceeded or a quota-style capacity code. Retry short rate pressure with Retry-After and backoff. For capacity codes, reduce usage, wait for pressure to drop, or move to a better-fit package.
When do warnings become urgent?
Rate state is healthy below 80%, warning at 80%, urgent at 95%, and blocked at the configured limit. A request that would exceed capacity returns a 429 response.
What should I monitor before launch?
Monitor request volume, token volume, image usage, rate-limit state, 429 response codes, and estimated cost trends before sending public traffic.
Free tools
Try these before your first live request
Client-side helpers for migration, validation, quota planning, and 429 troubleshooting.
Migration Generator
Generate or transform OpenAI SDK snippets to unlimitedcodex base URL, keys, and scopes.
Request Validator
Validate chat, embeddings, and image JSON payloads against unlimitedcodex OpenAI-compatible rules.
429 Decoder
Paste a 429 response body and headers to classify rate-limit vs quota pressure and see retry guidance.
Scope Configurator
Select workloads and output least-privilege scopes, .env.example, and a rotation checklist.
Rate Limit Planner
Plan backoff settings and quota monitoring checklist from expected RPS and team size.
Preparing for customer-facing traffic?
Pair rate-limit planning with error handling and webhooks.
Connect threshold monitoring, 429 behavior, package fit, and rate-state reviews before customers depend on the workflow.