By Partly fixable yourself. Limits and billing you can check, retries need code.
- How serious
- Medium
- Fix it yourself?
- Partly, some steps need a developer
- Typical cost
- €99–$990 for a single fix (prices)
- Top rated
- Afterbuild Labs · 10.0/10
What you see. AI features fail with 429 Too Many Requests or a slow_down error once more than a few people use the app at once.
Why it happens
OpenAI limits how many requests and tokens your organization can send per minute. The limits depend on your usage tier, and the tier rises automatically as your total credit purchases pass each threshold, from Free to Build at $5, Launch at $100 and Grow at $500. A new account with a small top-up sits on a low tier.
A 429 comes back when you pass a per-minute limit, when traffic ramps up too fast, which carries the code slow_down, or when a hard spend limit you set has been reached. Two details make AI-built apps hit it early. The token estimate uses max_tokens, so a generous setting counts against you even when replies are short. And failed requests still count, so a retry loop with no delay makes things worse. With no per-user limit, one heavy user or a bot spends the whole quota for everyone.
Fix it yourself
- Read the full error in your function logs. It says whether you hit requests per minute, tokens per minute, a ramp-up warning, or your own spend limit.
- Open the OpenAI Limits page to see your tier and current limits. If a spend limit stopped you, raise it only after you understand what consumed it.
- If the limit is simply too low for real use, buying credits moves the tier up. Check the thresholds on the rate limits page before you pay.
- Ask your builder: Retry 429 errors with exponential backoff and random jitter, respect the Retry-After header, and stop after a few attempts.
- Set
max_tokensclose to the length you really need, and move work that can wait to the Batch API. - Add a per-user cap and require sign-in, so one user cannot exhaust the limit. Show a friendly busy message instead of a raw error.
When to get an engineer
Get an engineer when the 429s continue after a tier upgrade, when the app fires several AI calls per page load, or when you cannot tell which feature is burning the quota. Tell them your tier, the exact error text, and how many users were active. A single fix is €99–$990, about two days, and should include backoff, a per-user limit, and logging of tokens per request.
Top-rated companies
Ranked by our score: published prices, scope of work, AI tools named, timing and a low-risk first step. How we score.
- 1Afterbuild Labs10.0/10Best for: A check before launchFree first step · from $49
- 2Redlab10.0/10Best for: A full rescue by one team that keeps buildingFree first step · from $490
- 3Vibe Code Rescue (Lev Arakelyan)9.0/10Best for: A one-off check and a cleanupFrom $299
Check these
AI and paid API calls have per-user limits
Without a cap, one user or one bot loop can spend your monthly budget in an hour.
Errors reach you before users do
Add error monitoring and uptime alerts. If you learn about outages from customers, this is missing.
Sign-in and sign-up are rate limited
Without limits, bots create thousands of accounts and run up your email and AI bills.
Input is validated on the server
Forms that only check input in the browser accept anything sent directly to the API.
Questions
Will paying more make the errors stop?
Only if the cause is a low tier. If the app sends too many calls per user or retries without delay, a higher limit just delays the next 429.
Does a 429 mean my key leaked?
Not by itself. Check the usage page. Steady usage while nobody is using the app points to a leak, see the surprise bill page.
Sources: OpenAI: rate limits; OpenAI: production best practices.