---
title: "Draft Models, Target Models & Parallel Verification"
description: "Where draft tokens come from, the acceptance rule at temperature zero and when sampling, accepted and bonus tokens, and why verification is lossless."
canonical_url: "https://fanout.sh/inference-eng/curriculum/16-02"
md_url: "https://fanout.sh/inference-eng/curriculum/16-02.md"
access: "pro-preview"
---

# Draft Models, Target Models & Parallel Verification

Where draft tokens come from, the acceptance rule at temperature zero and when sampling, accepted and bonus tokens, and why verification is lossless.

- Access: Fanout Pro (public preview)

## Public preview

The sequential decode lesson left a decode step able to score several positions for roughly the price of one, and left two questions with it. The positions have to come from somewhere, and whatever produces them will sometimes be wrong.

## In this lesson

- Where the candidates come from
- The rule at temperature zero
- The rule when the model samples
- Accepted, recovered, and bonus
- What lossless means

## Continue learning

Pro unlocks the complete lesson, video, practice, and code where available.

---
This representation contains public Fanout content only. Protected Pro lessons, account data, billing, checkout, and pricing are not included.

Browse the public content map: https://fanout.sh/sitemap.md