OpenRouter Publishes Guide to Fixing LLM API 429 Rate Limit Errors

OpenRouter published a guide on handling 429 Too Many Requests errors from OpenAI, Anthropic and its own API, testing three fixes on a 60-prompt batch. In its test, the OpenAI Python SDK's default retries left 23 of 60 requests failed in 12 seconds, while a retry that waits for Retry-After or X-RateLimit-Reset completed all 60 in 89 to 92 seconds in Python and 106 to 218 seconds in TypeScript. A model fallback finished all 60 in 13 seconds for $0.000066, and with provider failover on, all 150 pinned-provider requests succeeded where up to 32 failed with fallbacks off. The guide's own setup notes apply to OpenRouter accounts: running all 60 prompts requires at least 60 free-model requests left that day, since accounts that have not bought 10 credits get 50 free-model requests a day and 1,000 after that. The 20-per-minute cap applies only to :free models, and with BYOK, requests using your own provider key count against your limits at that provider.

Generated from reports · updated 12 h ago

Not enough continuous observations yet to draw a trend.