Back to glossary

Context window

Bağlam penceresi

D3

A context window is the finite amount of input context a particular model interface can use for a request or conversation state.

Review status: 2026-11-26

Technical explanation

Its available capacity is shared by the system instruction, accumulated conversation history, and current input.

Conceptual boundaries

A context window is a capacity limit, and input, generated output, and reasoning tokens can share its available budget in the cited Azure interface.

Provider-neutral example

A team can shorten older conversation turns before adding selected documents so the current request fits the model’s available context window.

Limitations

In the cited Microsoft API, the application must manage this finite capacity because the API does not automatically truncate or summarize conversation history.

Related concepts

Atomic claims and evidence

  1. 1.1Microsoft Learn, Best practices for the Phi Silica LanguageModel API
    Source
    Microsoft Learn, Best practices for the Phi Silica LanguageModel API
    Source role
    Authoritative source
    Exact locator
    Managing Context Length: finite shared context window for system prompt, history, and current prompt
    Supported claim
    Microsoft documents that a language-model context has a finite window shared by the system prompt, accumulated conversation history, and the current prompt.
    Last verification
    Review due
    Scope limitation
    This is product documentation for a Microsoft language-model API and does not prescribe a universal context-window implementation.
  2. 2.1Microsoft Learn, Foundry Models sold by Azure — model token limits
    Source
    Microsoft Learn, Foundry Models sold by Azure — model token limits
    Source role
    Authoritative source
    Exact locator
    Understand model token limits: input, generated output, and reasoning tokens share the available context budget
    Supported claim
    Azure Foundry documents a shared context budget in which input, generated output, and reasoning tokens use the available capacity.
    Last verification
    Review due
    Scope limitation
    This is provider-specific Azure Foundry documentation and does not prescribe a universal model limit.