OpenAI
- 429 insufficient_quota: usually your account's limits or billing
- 429 rate_limit_exceeded: usually your account's limits or billing
- 500 server_error: usually OpenAI's side
- 503 engine_overloaded: usually OpenAI's side
Anthropic
- 529 overloaded_error: usually Anthropic's side
- 429 rate_limit_error: usually your account's limits or billing
Gemini
- 503 UNAVAILABLE: usually Gemini's side
- 429 RESOURCE_EXHAUSTED: usually your account's limits or billing
How to read a code
A 429 is almost always a limit on your own account: requests or tokens per minute, or for OpenAI's insufficient_quota, money. A 500, 503 or Anthropic's 529 is the provider's own failure or load. Our checks send a few small requests to each API from several regions, so they show whether the API answers at all; they cannot see your account or load on one model. Each code's page says what that means for the reading you see there.
See also: how our checks work.