Early-Signal Detector

phenomenon · first seen 5 Oct 2026

time to first token

Also: TTFT

The latency from submitting a request to receiving the first generated token, a key metric for LLM serving.

Mentions per week

0119 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Earliest sources

Latest sources

Mentioned together with

KV cache reuseselective recomputationtokens per second