Unverified50% confidenceDecision RuleExact time
对流式响应场景，应重点关注首Token延迟（TTFT）而非整体完成时间
1
Sources
50%
Confidence
Long-term
Relevance
7/20/2026
First Seen
Sources
Related Claims
UnverifiedInference latency is typically decomposed into two core metrics: time to first token (TTFT) and subsequent token generation throughput.81% similarUnverifiedIf the timeout threshold for streaming responses is not configured properly, the client may terminate the response stream early due to a timeout judgment, and GLM is more prone to triggering timeouts when inference chain latency is longer.70% similarUnverified传统混合部署中Prefill阶段的大批量计算会阻塞Decode阶段的token生成，导致TTFT波动剧烈67% similarUnverifiedThe Traveling Salesman Problem with Time Windows (TSPTW) is a classic combinatorial optimization problem involved in itinerary planning.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/565595API
curl https://kongchang.com/api/v1/knowledge/claims/565595MCP
get_claim(id=565595)