Skip to the document
Madhuopen lab
Outcome School · Transformers and architecture10 min read

by Amit Shekhar · 29 March 2026

Paged Attention in LLMs

Paged Attention, a technique that solves the memory waste problem of KV Cache, allowing LLMs to serve many more users at the same time.

1,975 words#llm#ai#machine-learning4 recall cards

Paged Attention in LLMs
Read on Outcome School ↗then come back to lock it in
Before you read, guess

What memory allocation is required when a user initiates an LLM request?

Ten seconds, a guess, then read — a wrong guess still makes the answer stick.

What this article covers

  1. Quick Recap: KV Cache
  2. The Problem: Memory Waste in KV Cache
  3. What is Paged Attention?
  4. How Paged Attention Works
  5. Why Paged Attention Is So Effective
  6. Memory Sharing Across Requests

The article lives on outcomeschool.com. Read it there, then come back: the tutor in the margin has read it and will answer questions, and the questions below check what stayed.

Before you go

In one sentence, what was this chapter about?

From memory, without scrolling up. Writing it is what makes it yours; the grade is only to show you what you had.

How sure?