ModelRefs / Draft Model — AI Glossary
Draft Model — AI Glossary
A small, fast model used in speculative decoding to propose candidate tokens that the large target model then verifies.
Overview
Speculative decoding (Leviathan et al. 2023) pairs a draft model (~1–3B) with a large oracle model. The draft generates k tokens speculatively; the oracle verifies all k in one forward pass. Accepted tokens are kept; the first rejected token triggers a correction. Net effect: 2–4× throughput with identical output distribution.
Reference details
| Topic | inference |
|---|---|
| Also known as | speculative model, drafter |
| Last reviewed | 2026-06-24 |
Related terms
Primary source
Continue your research
Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to Draft Model — AI Glossary.
Frequently asked questions
What is Draft Model?
A small, fast model used in speculative decoding to propose candidate tokens that the large target model then verifies.
Is Draft Model the same as speculative model?
Yes — speculative model, drafter are common aliases for Draft Model.
What concepts are related to Draft Model?
Closely related concepts include speculative decoding, throughput, time to first token.