What is a context window in LLMs?

The context window is the maximum number of tokens (input + output) a model can process in a single request. Claude Opus 4.8 has a 1M token window, GPT-4o has 128K. Exceeding it causes errors or truncation.

How should I allocate my context window?

A typical RAG app allocates: 2-5K for system prompt, 10-30K for retrieved documents, 10-50K for conversation history, and 4-8K reserved for output. Use this planner to visualize exactly how much each section consumes.

Context Window Planner

Visualize & plan token allocation for your context window

31K / 1.1M3%

1.0M free

System Prompt

RAG / Documents

Conversation History

Reserved Output

Fits GPT-5.5

$0.2579/req~$7736/mo @1K/day

tok

Related:Token Counter·LLM Pricing