Standard Compute

All fixes · OpenAI · 429 · Rate limits

OpenAI “Rate limit reached for requests” (429)

Rate limit reached for requests

You’re sending more requests per minute (RPM) than your OpenAI tier allows. OpenAI returns HTTP 429 and rejects the request until the per-minute window resets. It’s a throughput cap, not a billing problem.

What causes it

  • Bursts of requests (parallel calls, retries, or an agent loop) exceed your tier’s requests-per-minute (RPM) limit.
  • Low usage tier: new and low-spend accounts get small RPM limits that rise as you spend more.
  • Background activity you don’t see (agent heartbeats, polling, tool calls) quietly consuming the per-minute budget.

How to fix it

  1. Add exponential backoff with jitter and retry on 429. Respect the Retry-After header if present.
  2. Throttle concurrency: cap how many requests you fire in parallel rather than bursting all at once.
  3. Raise your limit by moving up an OpenAI usage tier (Build, Launch and Grow unlock at $5, $100 and $500 in total credit purchases), or request a limit increase.
  4. Batch or queue non-urgent work so it spreads across the minute instead of spiking.

Running an agent? Agents are the classic cause: heartbeats, retries, and parallel tool calls add up fast and burn the RPM budget before you’ve sent a single message.

Your tier's limits

OpenAI rate limits by tier →
Requests and tokens per minute for each usage tier, and how to move up one.

Keep working on a fixed monthly plan

Standard Compute has none of the 5-hour or weekly windows other flat plans impose. Each key has a request ceiling of 300 to 2,000 requests a minute by plan and a fixed monthly compute budget, and transient upstream 429s are retried with fallback to another provider for the same model. A client that bursts past its key’s ceiling still gets a 429, so keep backoff in place.

Get your API key How it connects

Plans from $19 a month with $19 of compute. Eligible new accounts can try $0.25 of compute first, no card. Trial terms

FAQ

Is “rate limit reached for requests” a billing error?

No. It’s a per-minute throughput cap (RPM), not a quota or billing problem. You can have plenty of credit and still hit it. “You exceeded your current quota” is the billing one.

How long does an OpenAI rate limit last?

The per-minute window resets continuously, so a 429 usually clears within seconds to a minute. Back off and retry rather than hammering the endpoint.

Related errors