OpenAI Error 429: You exceeded your current quota

Your app or script calls the OpenAI API and gets a quota error, sometimes in this older form:

{
  "error": {
    "message": "You exceeded your current quota, please check your plan and billing details.",
    "type": "insufficient_quota",
    "code": "insufficient_quota"
  }
}

with HTTP status 429. Because 429 usually means "slow down", people add retries and wait — which never works here.

insufficient_quota indicates a billing or usage-limit problem, rather than a request-rate limit. Current responses can keep that broad error.type while giving a more specific error.code: credit_balance_exhausted, project_spend_limit_exceeded, organization_spend_limit_exceeded, or organization_usage_limit_exceeded. Read the code to find the cause. Retrying won't restore access until the credits or applicable limit change. (OpenAI error-code documentation)

Why it happens

1. No credit on the API account

The OpenAI API is billed separately from ChatGPT. A ChatGPT Plus or Pro subscription does not include API usage. If your API organization uses prepaid billing and its credits run out, requests fail with credit_balance_exhausted. Other API billing arrangements exist, so check your organization's billing setup.

Fix: go to the OpenAI platform billing page, add a payment method and buy credits. It can take a few minutes to take effect.

2. Credits expired

Prepaid credits and free trial grants can expire. Check the billing page for the expiry date.

3. A budget or spend limit was hit

Organization and project hard spend limits can stop API traffic even when credit remains. A spend alert only notifies you; it doesn't stop requests. For project_spend_limit_exceeded or organization_spend_limit_exceeded, raise or remove the reached hard limit if traffic should resume before the monthly reset. For organization_usage_limit_exceeded, request a higher OpenAI-assigned usage limit. (OpenAI spend-limit guide)

4. The key belongs to a different project or organisation

If you belong to several organisations or projects, the key in your .env might belong to one with no credit. Create a key in the project that has billing set up, and update your environment variable. (Environment Variables Explained)

5. The account is restricted

Rarely, accounts are limited for policy or payment problems. The billing page or an email will usually say so.

Real rate limits look different

A genuine rate limit returns 429 with a message like "Rate limit reached for requests" or "...tokens per min", and goes away if you wait and retry with backoff. insufficient_quota doesn't. Treat them differently in code — retry rate limits, but alert a human on quota errors. (429 Too Many Requests)

const billingCodes = new Set([
  'insufficient_quota', // older responses
  'credit_balance_exhausted',
  'project_spend_limit_exceeded',
  'organization_spend_limit_exceeded',
  'organization_usage_limit_exceeded',
  'usage_limit_exceeded',
])

if (err.status === 429 && billingCodes.has(err.code)) {
  // don't retry — alert with the code and show a friendly message
}

Stop it happening in production

Turn on auto-recharge so the balance tops up before it hits zero.
Set budget alerts that email you at, say, 50% and 80%.
Rate-limit your own users, so a bot can't burn through your credit overnight. (Stop Bots Running Up Your AI Bill)
Keep the key on the server. A key exposed in frontend code will be found and used by others. (Hide API Keys in an AI-Built App)
Cache reusable answers to repeated questions. Prompt caching reduces the cost of repeated input; it doesn't reuse the generated answer. (Caching Explained, Prompt Caching Explained)

The same thing with other providers

Anthropic's API says "Your credit balance is too low to access the Anthropic API" (a 400 error) in the same situation. OpenRouter and others have their own credit errors. The fix is always billing, not retries. (OpenAI API vs Claude API)

The summary

Inspect error.code: quota failures can mean exhausted credits or a hard spend/usage limit.
ChatGPT subscriptions don't cover the API; add credits on the platform billing page.
Check project budgets and which project your key belongs to; alert on this error instead of retrying.

EasySpawn gives your AI app a server-side backend where API keys stay secret and usage can be rate-limited per user — with Claude Code to wire in the guard rails. See how it works or join the waitlist.

"You Exceeded Your Current Quota" (OpenAI 429): What It Means and How to Fix It