Rate Limit Calculator

A token bucket is the most common rate limiting algorithm. The bucket holds a maximum number of tokens (burst size). Tokens are added at a fixed rate (RPM / 60 = tokens/sec). Each request consumes one token. When the bucket is empty, requests are rejected or queued. This allows short bursts while enforcing a long-term rate limit.

RPM (requests per minute) and RPS (requests per second) both measure API throughput. RPM = RPS × 60. Most rate limit docs use RPM (e.g., OpenAI: 10,000 RPM). When calculating delays between requests, convert to per-second: delay = 1000ms / RPS.

Divide your RPM by the number of concurrent workers. Each worker should wait at least (60,000ms / RPM) × workers between requests. Use a shared queue or Redis-based rate limiter (e.g., bottleneck, p-throttle) so workers coordinate instead of each maintaining their own timer.

Burst size is the maximum number of requests you can send instantly before the rate limit kicks in. For a 60 RPM limit with burst=10, you can send 10 requests immediately, then must wait ~1 second per additional request. Larger burst sizes are useful for handling traffic spikes without queuing.

If you fully exhaust a token bucket (burst size = N), the refill time is N × (60,000ms / RPM). For example, a 60 RPM limit with 10-token burst: refill = 10 × 1000ms = 10 seconds. The bucket is fully replenished after 10 seconds of no requests.