---
title: "Automatic Prefix Caching"
description: "How vLLM finds a reusable prefix: block identity, the lookup at admission, the hit counters and log field, and why a freed block can still be found."
canonical_url: "https://fanout.sh/inference-eng/curriculum/10-05"
md_url: "https://fanout.sh/inference-eng/curriculum/10-05.md"
access: "pro-preview"
---

# Automatic Prefix Caching

How vLLM finds a reusable prefix: block identity, the lookup at admission, the hit counters and log field, and why a freed block can still be found.

- Access: Fanout Pro (public preview)

## Public preview

Prefix reuse is the engine's default, and the repeated prefix computation lesson closed on two readings of it that make no sense next to each other. Three requests sharing almost all of their text queried the cache for 48 tokens and hit on none of them. Two other requests sharing a run of 266 tokens were answered with 256, not…

## In this lesson

- A block's identity
- The lookup at admission
- The counters and the log field
- One request's view of the cache
- A free block that is still findable

## 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