Unverified50% confidenceFactExact time
API限流通常以RPM（每分钟请求数）和TPM（每分钟Token数）两个维度执行，超限时返回429状态码并附带Retry-After响应头
1
Sources
50%
Confidence
Long-term
Relevance
7/12/2026
First Seen
Sources
多模型分层部署：AI网关值得引入吗？
redditr/mlops7/11/2026
Related Claims
VerifiedAI API的速率限制通常以RPM（每分钟请求数）和TPM（每分钟Token数）两个维度进行约束73% similarUnverifiedAnthropic对Claude API设有TPM（每分钟Token数）和RPM（每分钟请求数）上限73% similarUnverifiedAnthropic's API Rate Limiting mechanism measures request frequency in RPM (Requests Per Minute) and TPM (Tokens Per Minute).70% similarUnverifiedOpenAI按TPM（Tokens Per Minute）和RPM（Requests Per Minute）双重限制进行速率控制67% similarVerified供应商的速率限制通常在每分钟请求数（RPM）、每分钟 Token 数（TPM）、每天 Token 数（TPD）三个维度同时生效66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/492468API
curl https://kongchang.com/api/v1/knowledge/claims/492468MCP
get_claim(id=492468)