Draft model
A draft model is a language model that proposes candidate tokens for a target model to verify during speculative decoding.
Domain-Specific Small Language Models
The smaller, faster model in speculative decoding that proposes candidate tokens for verification by a target model.
- How it is used
- Proposes candidate tokens for a target model to verify during speculative decoding.
- How deep it goes
- Defines the proposal and verification roles, selection criteria, and acceptance tradeoffs without detailing decoding internals.
- Examples
- Pairs Llama 3.2 1B Instruct with Llama 3.2 3B Instruct as draft and target models.
- What you can do after
- Configure a draft model in LM Studio by selecting it alongside a compatible target model.
Read how Domain-Specific Small Language Models explains draft model →
Rearchitecting LLMs
A draft model proposes candidate tokens for a target model to verify during speculative decoding.
- How it is used
- Speeds target-model generation by proposing candidate tokens for parallel verification during speculative decoding.
- How deep it goes
- Explains the propose–verify loop and greedy-output equivalence, but does not detail model internals.
- Examples
- Uses predictable agent syntax and a five-token ZIP-code illustration to show drafting and correction.
- What you can do after
- Apply speculative decoding by pairing a draft model with a target model that verifies candidate tokens.
Also covered in
- Build a DeepSeek Model (From Scratch)In early access (MEAP)
What to know first
- language modelA language model is a computational model that represents patterns in linguistic sequences to assign probabilities, predict, generate, or evaluate language.
- generationGeneration is the production of an output by a computational model from supplied input or context, often using patterns learned from data.
Related concepts
- speculative decodingused by · Speculative decoding generates candidate tokens with a faster proposal process and verifies them with a target model before accepting them.
- MTP modulealternative to · An MTP module is a neural-network component that predicts an additional future token from a representation produced at the preceding prediction depth.
- target modelcontrasts with · A target model is the model that verifies candidate tokens proposed by a draft model during speculative decoding.
Questions about draft model
What is draft model?
A draft model is a language model that proposes candidate tokens for a target model to verify during speculative decoding.
Which Manning books explain draft model?
Domain-Specific Small Language Models and Rearchitecting LLMs.
What should I know before learning draft model?
Manning books that teach draft model usually introduce language model and generation first.
What is the difference between draft model and MTP module?
A draft model is a language model that proposes candidate tokens for a target model to verify during speculative decoding. An MTP module is a neural-network component that predicts an additional future token from a representation produced at the preceding prediction depth.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.

