Unverified60% confidenceFactTime unknown
对于500 Token的回答，流式输出可以让用户在首Token延迟（约500ms）后就开始看到内容，而非等待完整生成（约10秒）
1
Sources
60%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Related Claims
UnverifiedIt is generally considered that single-concurrency output above 100 tokens/second is sufficient to achieve a smooth typewriter effect, and at approximately 200 tokens/second, generating a 2,000-character long text takes only about 10 seconds.73% similarUnverifiedStreaming output can reduce the user-perceived time-to-first-token from several seconds down to the millisecond level.73% similarUnverifiedAnthropic's Prompt Caching API requires the cached prefix to reach at least 1,024 tokens, the cache TTL is typically 5 minutes, and it can be automatically renewed during active sessions.70% similarUnverifiedCopilot从停止输入到看到补全建议出现的整个链路通常在200-800毫秒之间完成69% similarUnverifiedOverly conservative VAD settings may wait 400–800ms after the user pauses before determining that a sentence has ended, and this waiting time is added on top of the system's perceptible latency.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/22460API
curl https://kongchang.com/api/v1/knowledge/claims/22460MCP
get_claim(id=22460)