---
title: "Why Separate Prefill and Decode"
description: "A long prompt makes other users' tokens late. Find whether shared execution, an overloaded queue or another part of the path causes it before splitting."
canonical_url: "https://fanout.sh/inference-eng/curriculum/21-01"
md_url: "https://fanout.sh/inference-eng/curriculum/21-01.md"
access: "pro-preview"
---

# Why Separate Prefill and Decode

A long prompt makes other users' tokens late. Find whether shared execution, an overloaded queue or another part of the path causes it before splitting.

- Access: Fanout Pro (public preview)

## Public preview

A long prompt arrives while other users are receiving short answers. Their next tokens become late, even though the server still has enough memory. Before adding GPUs, determine whether the delay comes from shared execution, an overloaded queue, or another part of the request path.

## In this lesson

- Shared execution
- The request budget
- Equal resources
- A decision worth testing
- Practice: choose the comparison

## 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