What is Paged attention?
Paged attention is an inference-time concept in large language model systems. It is part of the vocabulary teams use when they compare models, design prompts or review an AI system.
Definition
Paged attention is an inference-time concept in large language model systems. It is part of the vocabulary teams use when they compare models, design prompts or review an AI system. The term is used in inference discussions, where precision matters: two teams can use the same word to mean slightly different things. In this glossary, paged attention is treated as a practical concept rather than a formal specification.
Why it matters
Inference terms describe how a trained model turns a prompt into output: how tokens are sampled, cached and streamed. They drive latency, cost and the consistency of responses in production.
Example
Suppose you are comparing o3 Max with another model for a production task. Paged attention is one of the factors behind the numbers you see: o3 Max lists a 2,000,000-token context window, costs $17.82 per million input tokens and scores 87.0 on MATH-500. Understanding the term helps you decide which of those differences actually matter for your workload.
How to evaluate it
- Check whether paged attention is defined the same way in the documentation you are reading.
- Ask which measurement or trade-off the term is describing before comparing numbers.
- Separate vendor claims from independently observed behaviour on your own data.
- Write down the definition your team uses so reviews and evaluations stay consistent.
Related terms
Related guides
FAQ
What does Paged attention mean?
Paged attention is an inference-time concept in large language model systems. It is part of the vocabulary teams use when they compare models, design prompts or review an AI system.
Why does Paged attention matter?
Inference terms describe how a trained model turns a prompt into output: how tokens are sampled, cached and streamed. They drive latency, cost and the consistency of responses in production.
How is Paged attention used in practice?
Suppose you are comparing o3 Max with another model for a production task. Paged attention is one of the factors behind the numbers you see: o3 Max lists a 2,000,000-token context window, costs $17.82 per million input tokens and scores 87.0 on MATH-500. Understanding the term helps you decide which of those differences actually matter for your workload.