Pass IndexThe State of AISign in

Time to first token

How long you wait before the answer starts appearing.

how it is measured · the vocabulary

TTFTlatency

It governs whether an interface feels alive, and it is largely independent of how fast the rest streams. A model that begins in 200ms and writes slowly reads better than one that thinks for four seconds and then bursts.

Nearby

ARC-AGIBenchmarkBenchmark contaminationEloEval setGPQAHumanEvalLLM as judge