Unverified65% confidenceFactTime unknown
通过中转访问海外模型的延迟通常为300-800ms
1
Sources
65%
Confidence
Medium-term (~90 days)
Relevance
6/3/2026
First Seen
Valid until: 9/1/2026
Sources
Claude Code模型配置与费用对比：省钱实战攻略
bilibili小伍不是码牛
Related Claims
Unverified国内用户访问OpenAI、Anthropic等海外大模型服务时，跨太平洋传输的基础延迟（RTT）通常在150-300毫秒之间73% similarUnverified访问主内存的延迟高达200-300个时钟周期71% similarUnverified本地执行的联动响应延迟通常低于300ms，云端往返执行延迟普遍500ms-2s，对灯光/开关等即时反馈场景，本地执行体验差异明显71% similarUnverified本地部署延迟由三段构成：faster-whisper在GPU上STT约100-300ms、7B参数LLM首Token约200-500ms、TTS合成约100-300ms69% similarUnverified小模型推理延迟约500ms，而向量匹配延迟仅30-100ms68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/40077API
curl https://kongchang.com/api/v1/knowledge/claims/40077MCP
get_claim(id=40077)