Pass IndexThe State of AISign in

Speculative decoding

A small model guesses the next several tokens and the large one checks them all at once, which is faster than generating them one at a time.

how it works · the vocabulary

draft model

Purely an implementation trick: the output is identical to what the large model would have produced. It is one reason the same model runs at different speeds and prices at different sellers.

Nearby

AgentAgent memoryAlignmentAutoregressiveBM25ChunkingCold startComputer use