IonBin
All pages
Search
Concept

Context window

The context window is how much material a model can consider in one request: your instructions, documents, the conversation so far and its own reply, measured in tokens. Current flagship API models from Anthropic and OpenAI list about 1M tokens. App plans can allow much less, so check the plan you use.

Verified Oct 18, 2026 Review by Apr 16, 2027 5 sources

What it is

The context window limits the material a model can process in a request, measured in tokens. Instructions, tool definitions, retrieved content and conversation messages use capacity. Output and reasoning budgets also matter; check the model’s separate limits.

It is useful to think of this as a working desk, rather than permanent memory. Material stored in a chat or project is not necessarily all present in the model’s current input.

How it works

Applications may summarize older messages, select passages from files or retrieve details when needed. Claude documents automatic context management and project retrieval. Uploading a document does not prove that every page was extracted or considered.

As of October 18, 2026:

  • Claude’s current API model table lists 1M tokens.
  • GPT-6 Astra lists 1,050,000 tokens, with a separate 128,000-token output maximum.
  • Claude paid chat limits depend on model, including 200K, 500K and 1M; Claude Code and Cowork have their own tables.
  • ChatGPT Instant lists total windows of 27K on Free, 54K on Go/Plus and 128K on Pro. Its usable input allowance is smaller.

Benefits

A larger window can reduce the need to select passages before a task. It can help compare related material, provided the application actually supplies it. Keep source references and ask for evidence so you can check the answer.

Limits and risks

API cost depends on tokens processed and the provider’s input, cache, output and long-context pricing. Unused capacity is not billed simply because it exists; app subscriptions use different limits.

The 2023 Lost in the Middle study found position-sensitive retrieval failures in the models and tasks it tested. That motivates testing your own long-document task; it is not a benchmark result for every current model.

Checked specs

Snapshot from official sources on Oct 18, 2026.

Claude API current listed models1M tokens [1]
GPT-6 Astra API1,050,000 tokens; output limit separate [2]
Claude paid chat200K, 500K or 1M depending on model; Code/Cowork differ [4]
ChatGPT Instant total context27K Free, 54K Go/Plus, 128K Pro; input allowance is smaller [5]

Best-fit stacks

Agent compatibility

Tool in this entryHow agents connect
Claude Agent platform MCPAPI/CLIFile import
ChatGPT Agent platform API/CLIFile import
Gemini (Google) Agent platform API/CLIFile import

Levels come from each tool's recorded, sourced compatibility. How the levels work.

Sources

  1. Claude models overview · Anthropicofficial accessed Oct 18, 2026
  2. GPT-6 Astra model · OpenAIofficial accessed Oct 18, 2026
  3. Lost in the Middle · Liu et al.reference accessed Oct 18, 2026
  4. Paid Claude context windows · Anthropicofficial accessed Oct 18, 2026
  5. ChatGPT pricing · OpenAIofficial accessed Oct 18, 2026

Practitioner views are restated in our words and labeled as views, not measured results. Product facts come from official sources dated above.

Suggest an edit

IonBin's wiki isn't user-edited. Corrections go to our Directory Curator, who checks them against primary sources before anything changes. Spot something wrong or out of date on “Context window”? Email us with a link to a primary source.

Suggest an edit by email or write to contact@ionbin.com