OpenAI error 429 — "You exceeded your current quota"

Fix it yourself

Follow the repair

1. Read the code inside the error, not just the number. Log or print the full error response body, not only the status code. Buried in it is a code field, and that one word decides your entire fix:

   // The shape of the full error — find the "code" line in yours:
   { "error": { "type": "...", "code": "rate_limit_exceeded" } }
   //          rate_limit_exceeded = you are asking too fast
   //          insufficient_quota  = the account is out of money

2. For insufficient_quota: add billing to the API account. Open the OpenAI platform site, go to Settings, then Billing. Add a payment method or buy credits, and check that the monthly usage limit is not set to zero or already reached this month. The most common version of this error is a brand-new account that never had billing added at all. And to repeat it because it trips up so many people: paying for ChatGPT does not put a single cent into this account.
3. For rate_limit_exceeded: retry with backoff. When the limit really is speed, the fix is to wait and ask again — politely, with growing pauses. Start at one second, double the wait after each failure, and only retry 429s, not every error. A minimal version looks like this:

   async function callWithBackoff(fn) {
     let wait = 1000;                 // start with a 1-second pause
     for (let attempt = 1; attempt <= 5; attempt++) {
       try { return await fn(); }
       catch (err) {
         if (err.status !== 429) throw err;   // only retry 429s
         await new Promise(r => setTimeout(r, wait));
         wait *= 2;                           // 1s, 2s, 4s, 8s...
       }
     }
     throw new Error("Still rate limited after retries");
   }

4. Check your limits page before assuming you are capped forever. On the platform site, the Limits page shows your current requests-per-minute and tokens-per-minute for each model. New accounts start in the lowest tier, which is why a small burst can trip the limit during your first week. Tiers rise automatically as your payment history grows, so the numbers on that page — not a blog post — are the ones to plan around.
5. Slow the app down at the source. Backoff treats the symptom; this step treats the cause. Call the API once per user action instead of on every keystroke, page load, or re-render. Cache answers that repeat, batch small jobs into one request where you can, and put background work in a queue that sends at a steady pace instead of twenty requests in the same second.
6. If the very first call failed, it was never speed. Use this as your tie-breaker: a brand-new API key that receives a 429 on its first ever request cannot have exceeded a per-minute rate it never used. That is a billing wall announcing itself — no payment method, expired free-trial credits, or a monthly limit set to zero. Skip the retry code and go back to step 2.