Artificial Intelligence API rate limits

Based on the 918 providers in this industry that publish limits we have read. Providers can belong to several industries.

100median requests per minute, from 606 providers
30–360middle half of request rates

How much you can do

Requests per minute for 606 providers, using the limit that applies most broadly. Burst limits, daily and monthly quotas, and token or data limits are left out.

Under 1013% · 81 10 to 5917% · 104 60 to 29937% · 225 300 to 99918% · 110 1,000 to 9,99912% · 71 10,000 or more2% · 15

What the limit applies to

This decides whether another key adds capacity. Shares are of limit rows that name a scope.

API key or token35% · 307 Account or organization30% · 268 IP address19% · 165 Endpoint or operation7% · 64 User5% · 44 Global3% · 30 Model1% · 5

When you hit the limit

What providers document for clients and agents that need to back off. Of 918 providers with published limits.

16%say what response you get
28%name a Retry-After header
22%name a remaining-requests header

Limits this doesn't compare

211 providers also limit tokens, data, credits or concurrency, and 197 set daily or monthly quotas. 188 tie a limit to a named plan.

How APIs charge · Everything else about Artificial Intelligence →